Intelligent path planning algorithms for UAVs: Classification,complexity analysis,hybrid ablation insights,and future directions

Abstract

As unmanned aerial vehicle (UAV) technology has evolved, these systems are being increasingly utilized across diverse industries. However, controlling UAVs faces significant problems owing to several environmental circumstances and obstacles, making path planning a critical initial step for UAV operation. This paper offers an overview of UAV path planning research founded on intelligent algorithms, which are divided into three categories: computational intelligence (CI), machine learning (ML), and hybrid methods. Each category has been analyzed in depth to show its strengths, limits, and where it may be applied to UAV-related problems. The methodology includes a comparative analysis based on multiple performance metrics such as path length, flight time, collision avoidance, complexity, and environmental adaptability. Furthermore, the research covers the latest publications that deal with solving essential challenges of UAV path planning by using new hybrid algorithms and enhanced optimization methods. The results indicate that although each strategy offers specific strengths suited to particular scenarios, hybrid strategies are more likely to deliver greater flexibility and robustness, particularly in uncertain, and dynamic environments. These findings are significant for guiding future research in adaptive path planning and for supporting practical UAV applications such as autonomous delivery, aerial surveillance, disaster response, and environmental monitoring.

Keywords

unmanned aerial vehicles path planning machine learning computational intelligence performance metrics hybrid algorithms adaptive planning

Introduction

An unmanned aerial vehicle, often known as a drone, is a type of aircraft that does not require a human pilot on board. These vehicles can either be controlled remotely or autonomously by computers. Most recently, many fields have benefited from the use of UAVs. In agriculture, they have been employed for crop monitoring,¹ and irrigation management.² In the military field, they can provide many important functions such as the fulfilment of surveillance and reconnaissance,³ and combat operations.⁴ In building and engineering, the use of drones has helped improve project efficiency in safety monitoring.⁵ Moreover, drones are utilized in disaster management,⁶ delivery service,⁷ emergency medicine,⁸ and many other tasks. This makes them irreplaceable in a wide range of applications and an essential tool for various industries. Generally, UAVs operate in an obstructive and cluttered environment; for example, in an urban setting.⁹ Therefore, UAVs should rely on intelligent path planning algorithms to ensure safe and efficient flight paths. The principal objective of the path planning problem for UAVs is finding an optimal trajectory along which an unmanned aerial vehicle could travel from its starting location to some specified target while taking into account several factors such as terrain and obstacles.¹⁰

Path planning is a difficult task, and developing algorithms and techniques to address this issue has been the subject of numerous research papers. Over the last few years, some surveys have summarized the different methods. According to Table 1, they all contribute in some way to the advancement of comprehension and knowledge within a specific field of study. Actually, Zhao et al.¹¹ provided a detailed analysis of computational intelligence path planning algorithms in application for unmanned aerial vehicles. Debnath et al.¹² presented a survey of path planning algorithms in terms of energy efficiency, considering computation time, completeness, and path length. Cao¹³ reviewed some UAV path planning approaches classified into two categories based on their discovery order: traditional algorithms and intelligent algorithms. Luo et al.¹⁴ provided a comprehensive overview of UAV path planning approaches, which are divided into two classes based on algorithmic and functional levels. Ait Saadi et al.¹⁵ presented an overview of UAV path planning algorithms, grouped into five major categories, including classical methods, heuristics, meta-heuristics, machine learning, and hybrid techniques. This paper offers a comprehensive overview of UAV path planning research focused specifically on intelligent algorithms. It divides these approaches into three main categories: computational intelligence (CI), machine learning (ML), and hybrid methods.

Table 1.

Summary of review papers on UAV path planning.

Reference	Year	Summary
Zhao et al.¹¹	2018	This research examines the use of various computational intelligence algorithms in UAV path planning,the type of time domain (offline and online), and the type of environment models, including 2D and 3D.
Debnath et al.¹²	2019	The energy efficiency of every path planning method for a UAV is studied in this paper. Concurrently, the computation time, path length, and completeness are taken into account.
Ait Saadi et al.¹⁵	2022	This study explores UAV path planning methods, which are classified into five groups, including classical methods, heuristics, meta-heuristics, machine learning, and hybrid algorithms. Each category is analyzedfor its strengths and weaknesses.
Cao¹³	2023	This paper covers some UAV path planning algorithms and divides them into two categories based onthe discovery order of the algorithms (traditional and intelligent). Next, after listing the advantages of both intelligent and traditional algorithms, authors suggest using intelligent algorithms.
Luo et al.¹⁴	2024	In this research, path planning algorithms for unmanned aerial vehicles (UAVs) are presented and categorized according to algorithmic and functional levels. Furthermore, it thoroughly examines the benefits and drawbacks of every path planning algorithm, with the goal of analyzing their effectiveness.
This study	2025	It focuses on intelligent path planning algorithms and classifies them into 3 categories: computational intelligence (CI), machine learning (ML), and hybrid approaches. Improvements to computationalintelligence and machine learning algorithms are evaluated individually, with a critical analysishighlighting their advantages and disadvantages.

In Google Scholar, several key terms were used to gather research papers, including “intelligent algorithms for UAV path planning,”“machine learning algorithms for UAV path planning,” and “computational intelligence for UAV path planning.” This search strategy facilitated the emphasis of studies that only used intelligent algorithms for UAV path planning. Afterward, we looked at studies that proposed improved approaches specifically designed to address existing algorithms limitations by using specific terms such as “adaptive algorithms for UAV path planning,” and “hybrid algorithms for UAV path planning.” We also included searches on hybrid techniques, which explored combinations of intelligent algorithms as well as integrations of intelligent and classical methods, using terms like “hybrid algorithms for UAV path planning,” and “integrated CI and classical approaches.”

In this study, publications are collected from numerous databases, including IEEE, Springer, MDPI, Elsevier, and Wiley Online Library. Figure 1 presents the distribution of UAV Path Planning publications used in this study by databases. Due to the attention being paid to engineering and UAV-based technologies, the first place belongs to IEEE Xplore with almost 80 published articles. MDPI follows in second place with about 63 publications. Springer and Elsevier contribute equally. Beside them, Wiley Online Library and IOP Publishing provide a smaller number of papers.

Figure 1.

Distribution of UAV path planning publications by publisher database used in this study. A total of 213 publications were collected from six major databases: IEEE, Springer, Elsevier, MDPI, Wiley Online Library, and IOP Publishing.

The structure of this paper is organized as follows. Section 2 outlines key definitions, objectives, and constraints related to UAV path planning. Section 3 provides a succinct discussion of traditional algorithms used in UAV path planning, along with an analysis of their limitations. Section 4 classifies intelligent algorithms and discusses the various intelligent methods used in UAV path planning. Finally, the conclusion of this paper is presented in Section 5.

Key terms and definitions

The following section is intended to present some definitions and concepts that will be used in general to survey all research related to the path planning of UAVs using intelligent techniques.

Path planning

Path planning is the methodology of choosing an optimal route for UAV to move from a starting point to a target with consideration of constraints and obstacles.¹⁶

Path planning objectives

Collision avoidance

Collision avoidance is the ability of UAVs to identify obstacles and move safely without incurring any collisions or physical damage. UAV safety through obstacle avoidance is discussed in.^17,18

Path length

In their article,^19,20 authors focus on minimizing the distance traveled by UAVs from the starting point to the desired destination (path length).

Time-efficiency

Time-efficiency refers to the ability of UAVs to move from the initial point to the target point while avoiding obstacles in the least amount of time possible, as discussed by Darbari et al.²¹ and Huang et al.²²

Energy-efficiency

By definition, energy-efficiency is the minimum amount of energy required by UAVs to execute their missions, whether it be fuel, battery power, or any other type of energy. This concept has been explored by Modares et al.²³ and Fu et al.²⁴

Cost-efficiency

Cost efficiency refers to the total of various computational costs associated with UAVs, including hardware and software expenses, fuel costs, memory costs, battery charging costs, and CPU costs, as highlighted by Angley et al.²⁵ and Li et al.²⁶

Adaptability

Adaptability refers to the ability of a UAV to modify its planned path in real time in response to environmental changes. Liu et al.²⁷ and Ma and Chen²⁸ discuss adaptive path planning techniques for UAVs.

Path planning constraints

Altitude

Altitude is an essential element that needs to be balanced in order to ensure both safety and energy efficiency in UAV missions.

Energy consumption

During missions, energy is a major constraint for UAVs due to their limited battery or fuel capacity. This effectively reduces the capability of the UAVs to perform certain maneuvers.

Environmental constraints

Environmental constraints include weather conditions, such as high winds, rain, and snow; dynamic obstacles like moving vehicles and birds; and static obstacles, including buildings, trees, and any other kind of physical objects that could interfere with navigation.

Geometric constraints

The geometric constraints include factors that pertain to physical dimensions of UAVs, minimum turning radius, maximum/minimum velocity, and climb/descend angles.

Brief overview of traditional path planning algorithms

Traditional path planning methods have laid the foundation for the development of modern intelligent approaches. Conventional methods include, but are not limited to, grid-based methods,^29,30 potential-based methods,^31,32 and sampling-based methods.^33,34

Grid-based methods

In grid-based techniques, the environment is divided into cells, and algorithms such as A-star,³⁵ A-star and its variants,³⁶ improved A-star,^37,38 and Dijkstra’s algorithm,^39,40 are used to determine the optimal path from the start to the goal.

Graph-based methods

Different techniques like visibility graph,^41,42 Voronoi diagram,^43,44 depth-first search,⁴⁵ and breadth-first search⁴⁶ should be emphasized in graph-based path planning.

Sampling-based methods

Sampling-based methods randomly sample the points in the environment and use probabilistic frameworks to guide the search for the optimal path for the UAV, Examples of these techniques include Rapidly-Exploring Random Trees (RRT),⁴⁷ RRT*,⁴⁸ and Probabilistic Roadmaps (PRM).⁴⁹

Potential-based methods

The potential field methods utilize a combination of both repulsive and attractive forces that can be employed to steer a UAV to its target while avoiding any obstacles along the way. The repulsive force guides the UAV away from obstacles, while the attractive force attracts it toward the target. The Artificial Potential Field (APF) algorithm is the most useful path planning method for UAVs.^50,51

Complexity and limitations of traditional path planning algorithms

Classical algorithms have greatly improved UAV navigation proficiency, but their computational complexity renders them ineffective in large-scale or dynamic applications. Big O notation is a way to describe how an algorithm’s performance (in terms of runtime, memory) scales as the size of its input grows, particularly focusing on the worst-case scenario. For instance, the A* algorithm⁵² has a time complexity of O (b^d). In this case, b stands for the branching factor and d is the depth of the optimal path. This exponential complexity makes it inefficient in environments that have high dimensions or are densely cluttered. Furthermore, its space complexity also grows exponentially since it stores all the nodes that have been generated in memory. As a result, memory usage grows very quickly with map size and search depth, posing serious scalability problems. Dijkstra’s algorithm has a worst-case time complexity of O (n²), where n is the graph’s number of vertices. This complexity can be minimized by employing better data structures. For instance, implementing the algorithm with a binary heap brings complexity down to approximately O ((n+m) log n), and with Fibonacci heap to O (m+n log n), where m is the number of edges. These improvements help the algorithm become much faster, particularly for graphs that are not so dense.⁵³ Depth-First Search (DFS) and Breadth-First Search (BFS) both share the same time complexity of O (n+m) and are thus perfect for unweighted graphs but not so good in weighted or complex situations.⁵⁴ Sampling-based motion planning algorithms such as (PRM) and (RRT) typically have time complexity of order O (n log n) and space complexity usually considered to be O (n) where n is the number of samples.⁵⁵ Potential field-based methods, such as the Artificial Potential Field (APF) algorithm, generally take O (n) time and space complexity, where n is the number of obstacles in the environment. Although they offer real-time performances and simplicity of implementation, they are generally prone to local minima and do not provide good performance in complicated environments.⁵⁶ To overcome such limitations, researchers have been more and more adopting intelligent algorithms, particularly those based on computational intelligence (CI), machine learning (ML), and hybrid solutions. Intelligent algorithms offer a flexible and robust framework for addressing problems characterized by real-time environmental changes, sensor noise, and high-dimensional state spaces. However, even though CI and ML techniques significantly enhance the decision-making capacity and adaptability of UAVs, they can also come with great computational costs, such as long training times in ML techniques. To overcome this and to take advantage of the strengths of intelligent and traditional methods, hybrid algorithms have been researched increasingly. These techniques integrate traditional planning algorithms with CI or ML components to improve computational efficiency and path planning reliability. For example, the RFA-Star technique⁵⁷ integrates the classical A* algorithm with a feature attention mechanism. The proposed algorithm achieves a time complexity of O (n log n), showing strong computational efficiency and robustness in environments with dense obstacles. On the other hand, the RRT+DQN algorithm⁵⁸ exhibits also a time complexity of O (n log n), demonstrating robust performance and efficient obstacle avoidance in dynamic situations.

Classification and analysis of intelligent path planning methods

Traditional algorithms are valued for their ease of implementation and ability to rapidly generate feasible solutions in simple, well-defined environments. However, their application in complex problems (such as real-world scenarios) is probably the more challenging part. Consequently, researchers are increasingly using more advanced approaches such as artificial intelligence (AI) and machine learning (ML) to keep up with the growing requirements of UAV path planning. Figure 2 illustrates the classification of these intelligent algorithms.

Figure 2.

Classification of intelligent algorithms for UAV path planning.

Mathematical definition and distinction between CI, ML, and hybrid approaches

Intelligent algorithms for UAV path planning are generally grouped into three major paradigms: Computational Intelligence (CI), Machine Learning (ML), and Hybrid approaches. The three paradigms vary in conceptual foundation, mathematical expression, and mode of interaction with data and the environment. Computational Intelligence (CI) is a broad field that uses various nature-inspired and heuristic techniques to discover solutions to challenging problems. However, it is important to note that the meta-heuristic optimization algorithms make up a major part of CI, and they can be classified into single-solution and population-based methods depending on the number of candidate solutions maintained at each iteration.^59,60 A number of algorithm types belong to the population-based category, such as Genetic Algorithms (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Artificial Bee Colony (ABC), and Differential Evolution (DE). On the other hand, several types of algorithms fall into the single-solution-based category, such as Simulated Annealing (SA), Tabu Search (TS), and others. As a population-based techniques, Particle Swarm Optimization (PSO)⁶¹ searches for an optimal solution through a group of particles (swarm), with each representing a candidate solution. Velocity and position vectors for each particle are continuously updated by tracking the particle’s own best position ( $P_{Best}$ ) and the global optimal position of all particles ( $G_{Best}$ ). The two vectors are updated according to the following equation:

{\begin{matrix} V_{i}^{t + 1} = ω . v_{i}^{t} + c_{1} . r_{1} . ({P_{B e s t}}_{i}^{t} - X_{i}^{t}) + c_{2} . r_{2} . ({G_{B e s t}}^{t} - X_{i}^{t}) \\ X_{i}^{t + 1} = X_{i}^{t} + V_{i}^{t + 1} \end{matrix}

Where, $i$ denotes the index of the particle in the swarm, $t$ denotes the iteration number, $ω$ is the inertia weight, $c_{1}$ and $c_{2}$ are acceleration coefficients, and $r_{1}, r_{2} \in$ [0, 1] are random numbers. In contrast, algorithms such as Simulated Annealing (SA)⁶² represent single-solution-based meta-heuristic inspired by the physical process of annealing in metallurgy. During this procedure, the acceptance probability function or the Metropolis criterion, determines whether to make a transition from a current solution i to a potential new solution j:

p = {\begin{matrix} 1 if f (j) < f (i) \\ e^{(\frac{- (f (j) - f (i))}{T})} if f (j) \geq f (i) \end{matrix}

Where, $p$ denotes the transition probability, and $T$ is the current temperature.

On the other hand, one of the most powerful features of machine learning is that it can learn and improve its performance continuously from the data. Reinforcement Learning (RL) has been among the most effective ML approaches because it directly aims at decision-making in changing environments by combining machine learning and optimal control. In the context of path planning, an agent learns optimal behavior by interacting with the environment. During training, it receives a state as input, selects an action, observes the next state, and is then rewarded or penalized based on the outcome of its action. A multi-episode training process allows the agent to learn to select actions in accordance with maximum rewards, even under uncertain or dynamic conditions.⁶³ Reinforcement learning mathematically aims to find a strategy, also known as a policy $π (a | s)$ that maximizes the expected reward. The environment is commonly considered as a Markov Decision Process (MDP), where the goal is to compute the optimal value function $V^{*} (s)$ that represents the maximum expected cumulative reward. This is formally defined by the Bellman optimality equation ⁶⁴:

V^{*} (s) = \max_{π} E [\sum_{t = 0}^{t = n} γ^{t} R (S_{t + 1}, A_{t}, S_{t} | π, s)]

Here, $γ$ denotes the discount factor, $R$ is the reward function, $S$ is the state space, and $A$ is the action space. Reward shaping, the most common term for what is often referred to as reward engineering, involves modifying the reward function $R$ to provide additional, informative feedback in order to accelerate the learning process, and push the agent toward desirable behavior. For example, in path planning with obstacle avoidance, the agent can be penalized not only for direct collisions but also for navigating too close to obstacles while flying. This approach promotes safer navigation behavior as well as the generation of more reliable paths. Potential-based is one of the most popular reward shaping techniques, where rewards are modified by adding a potential function $Φ (s)$ ⁶⁵:

R' (s, a, s') = R (s, a, s') + γ Φ (s') - Φ (s)

Here, $R^{'} (s, a, s^{'})$ is the modified reward function, $R (s, a, s^{'})$ is the reward function, $γ$ denotes the discount factor, and $Φ (s)$ is the potential function.

Hybrid approaches represent a third paradigm that combines elements of Computational Intelligence (CI), Machine Learning (ML), and sometimes traditional path planning techniques. For instance, In CI with traditional algorithms (TCi), methods like Genetic Algorithms (GA) are used for generating and optimizing candidate UAV paths, while A* is employed to refine those paths using heuristic-based search.⁶⁶ In ML with traditional algorithms (TMl), one of the usual hybridizations is the use of reinforcement learning coupled with APF technique for UAV path planning improvement in dynamic environments. RL is utilized as a global planner, searching for targets and planning optimal route, while APF is used as a local controller, giving real-time obstacle avoidance.⁶⁷ In CI with ML algorithms (CiM), researchers often combine Particle Swarm Optimization (PSO) with Deep Q-Networks (DQN). PSO provides global search capabilities for UAV path planning optimization, whereas DQN is focused on learning a policy, enabling the hybrid model to balance exploration, and exploitation strategies.⁶⁸

Performance metric selection and trade-offs

In this paper, several metrics are considered in order to evaluate and compare the performance of intelligent path planning algorithms for UAVs. These metrics include path length, execution time, flight time, collision avoidance, energy efficiency, robustness, stability, and convergence speed. However, selecting the right metric is highly dependent on UAV mission-specific requirements and operational constraints. For instance, path planning of multi-UAV in 3D heterogeneous environment with variable coverage of GNSS satellites is targeted toward the minimization of path length and flight time.⁶⁹ In contrast, for the case of environments where safety is of utmost priority, such as maneuvering through dynamic threat environments, performance metrics like collision avoidance and robustness are essential in order to ensure safe navigation and successful mission completion.⁷⁰ For long-duration UAV missions such as in precision agriculture, energy efficiency is a key performance metric. An example of this is a UAV-based irrigation system for olive farms.⁷¹ Additionally, in complex terrains such as bridge construction sites,⁷² researchers have been concerned with improving convergence speed and stability. Due to the different objectives of UAV operations, many recent studies have also focused on multi-objective optimization, where multiple metrics such as path length, execution time, energy efficiency, and stability are balanced simultaneously.^73,74 However, optimizing multiple criteria often involves inherent trade-offs, where Improving one metric may lead to the deterioration of another. These sorts of trade-offs need to be appreciated in selecting intelligent path planning algorithms suitable to the specific needs of UAV missions. Some of the most common trade-offs encountered in existing UAV path planning research are outlined below: 4.2.1 Path length versus Collision avoidance: In order to remain safe, a UAV can divert from the shortest route in order to avoid obstacles. This safety requirement may lead to longer paths, longer flying time, and therefore increased energy consumption. Conversely, when minimization of path length is preferred, the UAV might follow a more straight-line path that poses a greater chance of collision. 4.2.2 Energy Efficiency versus Computational time: Enhancing the energy efficiency of UAV path planning might require the application of intelligent algorithms such as reinforcement learning (RL) methods, which are capable of generating smooth paths. However, these approaches are then accompanied with large computational time, particularly in dynamic and uncertain environments where there is a requirement of real-time adaptation and learning. 4.2.3 Convergence speed versus Stability: Recent studies show that convergence speed and stability trade off fundamentally. Algorithms that converge faster tend to be less stable in dynamic environments, whereas those with emphasis on greater stability often experience slower convergence speeds, and require longer training times. This trade-off is intrinsic to many machine learning methods, including deep reinforcement learning.

Computational intelligence algorithms

Computational Intelligence (CI)⁷⁵ refers to a collection of computational methods and approaches inspired by natural processes. These methodologies are particularly effective in solving complex, real-world problems where traditional mathematical models or conventional techniques are not feasible. Several Computational Intelligence (CI) algorithms have been employed for UAV path planning, as summarized in Table 2. There are two major subsets within computational intelligence: Swarm Intelligence (SI) and Evolutionary Algorithms (EAs).

Table 2.

Summary of computational intelligence algorithms for UAV path planning.

Class	Algorithm	Year	Env	Benchmarktype	SingleUAV	Results or performances
Swarm intelligence (SI)	Improved PSO⁷⁶	2020	Dynamic	Synthetic	Yes	Robustness of improved PSO compared to c-PSO and w-PSO in terms of path generation and time efficiency.
	Modified PSO⁷⁷	2023	Dynamic	Synthetic	No	Effectiveness of modified PSO compared to state-of-the-art methods in terms of distance traveled, energy-efficiency, and travel time.
	IACA⁷⁸	2022	Static	Simulated (Matlab)	Yes	Efficiency of IACA compared to the IABC and GA in terms of path length, convergence speed and flight time.
	IABC⁷⁹	2022	Dynamic	Synthetic	Yes	Effectiveness of IABC algorithm compared to PSO, ABC, SABC and GABC in terms of stability, convergence speed, and path generation.
	NOGWO⁸⁰	2024	Dynamic	Simulated	Yes	NOGWO algorithm gives better performance compared to GWO, WOA, DE, ALO, and SCA in terms of convergence speed and precision.
Evolutionaryalgorithms	Improved GA⁸¹	2014	Static	Synthetic	Yes	Efficiency of the improved GA compared to traditional GA in terms of convergence rate, collision avoidance, and computation time.
	GPSA-GA⁸²	2022	Static	Real-world	Yes	In comparison with conventional GA, GPSA-GA gives better results in terms of path length and convergence speed.
	DE-DYTS⁸³	2024	Dynamic	Synthetic	No	Superiority of DE-DYTS compared to DE, IMODE, and SHADE in terms of path optimization and risk avoidance.
	DECSMU⁸⁴	2024	Dynamic	Simulated	No	Efficiency of DECSMU in solving the multi-UAV path planning problem
Other CIalgorithms	SJSA⁸⁵	2020	Dynamic	Synthetic	No	SJSA gives better performance compared to DIDE, MGA, DISA, and the exact algorithm CPLEX in terms of running time and convergence rate.
	Improved TS⁸⁶	2022	Static	Synthetic	Yes	Efficiency of variable neighborhood Tabu search Algorithm in solving large-scale combinatorial optimization problem
	MVO⁸⁷	2018	Dynamic	Simulated	Yes	Superiority of MVO algorithm compared to GWO, WOA, DA, ALO, and MFO in terms of convergence rate, execution time, and path cost.

Swarm intelligence

Today, Swarm Intelligence (SI)⁸⁸ is generally recognized as a branch of Computational Intelligence (CI) inspired by the collective behaviors of biological organisms like bees, ants, birds, fireflies, bats, monkeys, and wolves. Algorithms in this field include Particle Swarm Optimization (PSO),⁸⁹ Ant Colony Optimization,⁹⁰ Artificial Bee Colony (ABC),⁹¹ and Gray Wolf Optimizer (GWO).⁹²

Nayeem et al.⁷⁶ suggested an improved Particle Swarm Optimization (PSO) algorithm, called n-PSO which utilizes a dynamic neighborhood approach to enhance particle diversity. The effectiveness of n-PSO was tested in a 3D environment with static obstacles. Test results showed that the n-PSO algorithm gives better performance compared to constriction factor PSO (c-PSO)⁹³ and time-varying inertia weight PSO (w-PSO)⁹⁴ in terms of path generation and time cost.

Sonny et al.⁷⁷ proposed a modified PSO algorithm for solving the path planning problems related to drone delivery, aiming to provide the most energy-efficient route for UAVs from the starting point to the destination. The performance of PSO algorithm was assessed in three different scenarios featuring 12 static obstacles. Experimental results demonstrated that modified PSO outperforms state-of-the-art methods in terms of path length, energy consumption, and travel time.

Qi et al.⁷⁸ suggested an improved ant colony algorithm (IACA) for optimizing the UAV path planning. The performance of the IACA algorithm was assessed in a 2D simulation environment using a grid map specifically designed for multi-rotor UAV. In comparison with GA and basic ant colony algorithm (BACA), IACA gives better results in terms of path length, convergence speed and flight time.

In the work of Lin et al.,⁷⁹ an improved artificial bee colony algorithm based on multi-strategy synthesis (IABC) was proposed to optimize UAV path planning in a complex, urban environment. The effectiveness of the improved ABC was evaluated in a 3D complex environment across three cases, with four obstacles in the first case, six obstacles in the second, and 12 obstacles in the third. Simulation results proved the effectiveness of IABC algorithm compared to PSO, ABC, SABC,⁹⁵ and GABC⁹⁶ in terms of stability, convergence speed, and path generation.

Rao et al.⁸⁰ proposed an enhanced gray wolf optimizer, called, multi-strategy collaborative gray wolf optimization algorithm (NOGWO), which includes additional strategies like random walk strategy, opposition-based learning, and a novel convergent factor for solving the UAV path planning problem. Test results demonstrated the effectiveness of NOGWO algorithm compared to GWO, WOA,⁹⁷ DE,⁹⁸ ALO,⁹⁹ and SCA¹⁰⁰ in terms of convergence speed and precision. Additionally, NOGWO outperformed the three swarm intelligence algorithms: GWO, SCA and WOA in terms of path length and cost efficiency.

Evolutionary algorithms

Evolutionary algorithms¹⁰¹ are a class of Computational Intelligence methods that use concepts of natural selection and genetics to find solutions to complex optimization problems. Notable algorithms in UAV path planning include Genetic algorithms (GA),¹⁰² and differential evolution (DE).¹⁰³

Wang and Chen⁸¹ developed a novel approach using an improved genetic algorithm (GA) for solving the UAV path planning problem. In this paper, three types of path planning schemes are presented based on prior knowledge: Directly Through, Insertion Through, and Bypass Through. Simulation results showed that the prior knowledge based genetic algorithm outperforms the traditional GA in terms of convergence rate, collision avoidance, and path optimization.

Yuan et al.⁸² presented an improved coverage path planning algorithm based on the combination of the good point set technique with genetic algorithm (GPSA-GA) for a fixed-wing UAV. In order to assess its robustness, GPSA-GA was tested 20 times in coverage areas of 20, 50, 80, and 100 paths, respectively. Experimental results demonstrated the superiority of the GPSA-GA compared to conventional GA in terms of path length and convergence speed.

Yin et al.⁸³ suggested an adaptive differential evolution, called DE-dynamic Thompson sampling (DE-DYTS) based on the integration of dynamic Thompson sampling into DE to address the collaborative path planning problem for multiple UAVs. The performance of DE-DYTS algorithm was evaluated in a 3D environment based on three experiences using two UAVs in the first, three UAVs in the second, and five UAVs in the third. In comparison with DE, IMODE,¹⁰⁴ and SHADE,¹⁰⁵ DE-DYTS gives better results in terms of path optimization, and risk avoidance.

Zeng et al.⁸⁴ developed a differential evolution-Cooperative Search for Multiple UAVs (DECSMU) for solving the multi-UAV path planning problem. In this method, a dynamic reward function was developed to improve coverage efficiency and energy consumption and a distributed model predictive control (DMPC) strategy was employed to achieve the collaborative coverage of multiple UAVs. The effectiveness of DECSMU algorithm was validated in three different search regions: two convex regions (a circular region and a rectangular region) and one non-convex region, using a group of four UAVs.

Other CI algorithms

In addition to the CI algorithms discussed above, there are many other algorithms that can solve UAV path planning problems or help improve performance, such as simulated annealing (SA),¹⁰⁶ Tabu search (TS),¹⁰⁷ Multi-Verse Optimization (MVO).¹⁰⁸

Huo et al.⁸⁵ developed a novel simulated annealing algorithm, called Swap-and-Judge Simulated Annealing (SJSA) algorithm based on the integration of Swap-and-Judge strategy into SA for solving Multi-UAVs path planning problem in disaster relief. The performance of SJSA was evaluated in six scenarios with different numbers of target nodes and UAVs. Simulation results demonstrated the superiority of SJSA compared to DIDE, MGA, DISA, and the exact algorithm CPLEX in terms of running time and convergence rate.

Tong et al.⁸⁶ proposed an improved variable neighborhood Tabu search algorithm for solving the UAVs path optimization problem for truck–drone delivery. The experimental results proved that the proposed method outperformed traditional truck delivery in terms of delivery time. This approach effectively optimizes the truck–drone delivery route, improving overall efficiency.

In the work of Kumar et al.,⁸⁷ a Multi Verse Optimizer (MVO) algorithm was suggested for UAV path planning to improve the Quality of Service (QoS) in UAV environment. The robustness of the proposed algorithm was tested in 6 different cases. Experimental results showed that MVO algorithm gives better performance compared to GWO,¹⁰⁹ WOA,⁹⁷ DA,¹¹⁰ ALO,⁹⁹ and MFO¹¹¹ in terms of convergence rate, execution time, and path cost.

Critical analysis

In this section, various UAV path planning algorithms based on computational intelligence are examined as effective solutions for handling complex and dynamic environments and adapting to a wide range of UAV missions. On the one hand, these methods have their merits in some aspects; on the other hand, they are suboptimal since most of them converge more slowly and suffer from falling into local optima, as summarized in Table 3.

Table 3.

Advantages and disadvantages of computational intelligence algorithms.

Algorithm type	Algorithm	Advantages	Disadvantages	Applications
Swarmintelligence (SI)	PSO^112–114	- High robustness- Easy implementation- Rapid convergence in the early stage	- Easy to fall into local optima- Limited particle diversity- Slow convergence in the later stage	- Energy efficiency- Path optimization in both static and dynamic environments- 3D path planning
Swarmintelligence (SI)	ACO^115–117	- Fast convergence in the later stage- Strong adaptability	- Easy to fall into local optima Slow convergence in the initial stage	- Obstacle avoidance in complex environments- Optimal path construction
Swarmintelligence (SI)	ABC^118,119	- Simple implementation and few control parameters- High adaptability	- Slow convergence speed- Premature convergence	- Multi-objective optimization- Dynamic real-time route planning
Swarmintelligence (SI)	GWO^120–122	- Easy realization- Flexibility- Fast convergence speed	- Premature convergence- Slower convergence in high dimensional problems	- Suitable for solving a wide range of optimization problems
Evolutionaryalgorithms (EAs)	GA^123,124	- Strong global search ability	- Poor local search ability- Slow convergence- High computational costs	- Complex and dynamic environments- Multi-objective optimization
Evolutionaryalgorithms (EAs)	DE^125–128	- Few parameters and simple implementation- Fast convergence speed	- Premature convergence- Parameter sensitivity	- Multi-objective optimization- Strong performance in real-world optimization problems
Other CI algorithms	SA^129–131	- It avoids local minima- Strong exploitation capability	- Slow convergence speed	- Complex and dynamic environments- Suitable for solving a wide range of optimization problems
	TS⁸⁶	- Easy to implement- Strong local optimization capability	- Memory demands- Not appropriate for large-scale optimization	- Complex combinatorial problems
	MVO^132–134	- Easy to implement- Strong global optimization ability- It avoids local optima	- Limited accuracy in solution quality- Slow convergence in the later stages	- Multi-objective optimization

To address these limitations, recent advancements in computational intelligence (CI) have employed hybrid methods,^135,136 adaptive strategies,^137,138 and enhanced optimization techniques.^139,140 In hybrid methods, different algorithms are implemented together to take advantage of their respective strengths and compensate for their weaknesses. For example, combining PSO with a similar computational intelligence algorithm like GA,¹⁴¹ enhances the ability to escape local minima while benefiting from GA’s high-quality solutions. The integration of machine learning algorithms is also widely used, such as merging GA with Q-learning algorithm¹⁴² to leverage GA’s optimization capabilities and QL’s real-time decision-making. Additionally, adaptive techniques such as adaptive population size enable differential evolution to overcome population stagnation and enhance the algorithm’s search performance.¹⁴³ Finally, improved optimization techniques, such as introducing Levy and Cauchy distributions to enhance the employed bees’ search strategy,¹⁴⁴ and using mutation crossover factor optimization strategy in the Differential Evolution (DE) algorithm to overcome the problem of path planning of UAVs in complex environments.¹⁴⁵ Table 4 presents a collection of research that addresses challenges within CI algorithms, highlighting improved methods and techniques designed to enhance the ability to escape local optima, convergence speed, path generation, and adaptability for UAV path planning applications.

Table 4.

Recent improvements in computational intelligence algorithms for UAV path planning.

Algorithms	Limitations	Improved/hybridalgorithms	Solution methods
PSO	- Easy to fall into local optima and slow convergence	IPSO¹⁴⁶	- Introduces the improved chaotic-VAINDIW into PSO to enhance the convergence speed and avoid falling into a local optimal solution
PSO	- Easy to fall into local optima and slow convergence	MPSO¹⁴⁷	- Employs chaos initialization and robust mutation mechanism
ACO	- Easy to fall into local optima	MACO¹⁴⁸	- The metropolis criterion is introduced into the node screening mechanism of ACO algorithm
ACO	- Slow convergence	IACA⁷⁸	- Introduces a dynamic adjustments factor and an enhanced state transition
ABC	- Easy to fall into local optima and slow convergence	EABC¹⁴⁹	- Employs a novel search equation and selection mechanism to improve convergence speed and avoid premature of the algorithm
ABC	- Easy to fall into local optima and slow convergence	FC-ABC¹⁵⁰	- Uses an adaptive neighborhood search technique to improve the convergence speed and ε-Boltzmann selection strategy to avoid falling into local optima
GWO	- Premature convergence	GED-GWO¹²⁰	- Integrates GWO with Gaussian estimation of distribution (GED) to avoid premature convergence
GA	- Poor local search ability and slow convergence speed	IDPGA¹⁵¹	- Uses three offspring generation methods (inbreeding, crossbreeding, and random generation) to enhance local and global search capabilities and improve the convergence speed
GA	- Poor local search ability and slow convergence speed	ERRT-GA¹⁵²	- Combines improved GA with new RRT initialization algorithm to improve the convergence speed
DE	- Parameter sensitivity	DMPSADE¹⁵³	- Employs a self-adaptive DE algorithm with discrete mutation control parameters to enhance robustness
DE	- Premature convergence	Chaotic DE¹²⁸	- Combines DE with chaotic theory to avoid premature convergence of DE algorithm
SA	- Slow convergence speed	SA-PSO¹⁵⁴	- Integrates SA with PSO to avoid falling into local optima and improve the convergence speed
MVO	- Limited accuracy in solution quality	Modified MVO¹³³	- Introduces modifications to MVO to enhance the quality of the solutions

Machine learning algorithms

In order to solve the UAV path planning problem, a variety of machine learning techniques were used, as summarized in Table 5. These techniques can be classified into five main categories: supervised, unsupervised, reinforcement learning, neural network, and deep reinforcement learning.

Table 5.

Summary of machine learning algorithms for UAV path planning.

Classification	Algorithm	Year	Environment	Benchmarktype	Single UAV	Results or performances
Supervisedlearning	SVM¹⁵⁵	2014	Static	Synthetic	Yes	SVM provides good path optimization results.
	GP-FMS¹⁵⁶	2023	Static	Simulated	Yes	Efficiency of GP-FMS compared to the classical methods.
	DLR-KNN¹⁵⁷	2024	Static	Simulated	Yes	Robustness of DLR-KNN compared to K-means and one-class SVM in terms of detection rate and precision.
	PDE¹⁵⁸	2020	Dynamic	Simulated	No	Efficiency of PDE compared to MILP in terms of path length and time consumption
Unsupervisedlearning	Improved SOM¹⁵⁹	2020	Static	Simulated	Yes	Effectiveness of improved SOM in terms of path optimization
Unsupervisedlearning	Landmarkclustering¹⁶⁰	2018	Dynamic	Simulated (Matlab)	No	The algorithm generates a safe and smooth flight path
Reinforcementlearning	MAPPO¹⁶¹	2022	Dynamic	Simulated(Unreal Engine 4)	No	The efficiency of MAPPO compared to traditional reinforcement learning method
	Q-learning¹⁶²	2023	Dynamic	Simulated	No	Q-learning outperforms other path planning approaches in terms of time learning and path length.
	IQL¹⁶³	2018	Dynamic	Simulated	Yes	IQL achieves better results compared to classical Q learning in terms of learning efficiency, path length and search performance.
	Improved RL¹⁶⁴	2024	Dynamic	Simulated	Yes	Effectiveness of APPA-3D compared to: classical algorithms (APF, RRT, and A*) and Q-learning-based algorithms (DFQL, IQL, and MEAEO-RL) in terms of path length and time-efficiency.
Neuralnetwork	oHJB¹⁶⁵	2020	Static	Simulated	Yes	In comparison to other neural network algorithms, oHJB performs better in terms of path length, traveling time, and energy consumption.
	DCNN-GA¹⁶⁶	2021	Dynamic	Real-world	Yes	The robustness of DCNN-GA compared to other deep neural network models.
	DNN-Adamoptimization¹⁶⁷	2024	Dynamic	Simulated	Yes	The efficiency of DNN-Adam optimization compared with 11 state-of-the-art peer deep neural networks
Deepreinforcementlearning	Improved DDPG¹⁶⁸	2020	Dynamic	Simulated	Yes	The effectiveness of improved DDPG in terms of tracking accuracy and obstacle avoidance.
	DQN-LSTM¹⁶⁹	2018	Dynamic	Simulated	Yes	Simulation results showed that RPP-LSTM is practical and feasible.
	Deep SARSA¹⁷⁰	2018	Dynamic	Simulated	No	Deep-Sarsa gives good results in terms of safe path generation under dynamic environment
	DDPG¹⁷¹	2020	Dynamic	Simulated	Yes	The robustness of DDPG compared to traditional RL training scheme and RRT*
	Improved D3QN¹⁷²	2020	Dynamic	Simulated	Yes	Efficiency of improved D3QN algorithm compared to DDQN and DQN algorithms in terms of path optimization

Supervised learning

Chen et al.¹⁵⁵ introduced a novel method based on Support Vector Machine (SVM). In a three-dimensional complex environment, the robustness of the SVM model was tested. The outcomes of the simulation showed that SVM creates a safe and smooth flight path as it approaches obstacles.

Muñoz et al.¹⁵⁶ proposed an approach which combines Gaussian Process (GP) regression model and the Fast Marching Square (FMS) method to develop an informative path planning and exploration algorithm. Based on the conducted tests, the weights were assigned to each method, and then the mean absolute error (MAE) and predictive variance were calculated. Testing the proposed algorithm in a real environment demonstrated its effectiveness compared to traditional methods, such as the Boustrophedon approach, by providing a more accurate depiction of the environment.

Alos and Dahrouj¹⁵⁷ proposed a hybrid approach, called Dynamic Linear Regression with K-Nearst Neighbor (DLR-KNN) for optimizing UAV path planning by addressing contextual faults within the UAV system. Dynamic Linear Regression is used for estimating the values of a focused attribute, while K-Nearst Neighbor is employed to classify the estimation error values into “Normal” and “Abnormal” categories. Experimental results proved that DLR-KNN outperformed K-Means and One-Class SVM algorithms in terms of detection rate, accuracy, and false alarm rate.

Radmanesh et al.¹⁵⁸ proposed a decentralized approach using a partial differential equation (PDE) to create collision-free 3D routes for multiple UAVs operating in common airspace. The suggested approach was compared with centralized and sequential planning techniques. Test findings showed that the partial differential equation (PDE) outperformed Mixed Linear Programing (MILP) in terms of path length and execution time.

Unsupervised learning

Pierre et al.¹⁵⁹ suggested an enhanced Kohenen’s Self-Organizing Map (SOM) approach using a combination of competitive learning with particle physics to develop an optimal trade-off solution for UAV multi-objective path planning. Three scenarios were used in a 3D environment to evaluate the performance of the enhanced SOM method. The outcomes showed that the proposed SOM can effectively avoid obstacles and escape falling into local minima.

Tartaglione and Ariola¹⁶⁰ suggested a novel decentralized path planning method based on the leader-follower technique and Quality Threshold Clustering (QT clustering) algorithm to generate collision-free 2D paths for multiple UAVs working in unknown environment. By simulating different scenarios, the results demonstrated the effectiveness of the proposed technique, as the UAVs successfully completed the flight mission while avoiding collisions with obstacles throughout their trajectory.

Reinforcement learning

Chen et al.¹⁶¹ suggested multi-agent reinforcement learning for optimizing multi-UAV path planning using a centralized training–decentralized architecture (CTDE). A long short-term memory (LSTM) layer was added to the proximal policy optimization (PPO). The experimental findings showed that the proposed algorithm significantly enhanced the performance and overcame the problem of making-decision with incomplete information.

Sonny et al.¹⁶² suggested a Q-learning algorithm to efficiently plan the path of UAVs in an environment containing both static and dynamic obstacles. The performance of the proposed approach was evaluated in terms of learning time and path length. The results showed that the proposed algorithm minimizes the total distance traveled by the UAV when compared to other approaches.

Yan and Xiang.¹⁶³ proposed an improved Q learning algorithm (IQL) for solving the UAV path planning problem in unknown environment. This paper introduces a new action selection strategy that combines the ε-greedy strategy with the Boltzmann strategy, as well as a novel Q-function initialization method, to enhance the algorithm’s performance. The simulation results proved that IQL algorithm outperformed the classical Q learning algorithm in terms of learning efficiency, path length and search performance.

Wang et al.¹⁶⁴ proposed an improved reinforcement learning algorithm to enhance the optimization of UAV path planning for unknown complex 3D environments. Three anti-collision avoidance strategies experiments, including a fight path opposing collision, a pursuit collision, and a cross collision, were conducted to evaluate the efficacy of the developed APPA-3D algorithm. The results validate the APPA-3D algorithm’s effectiveness in terms of collision avoidance. As an additional experiment, authors conducted a comparative analysis between APPA-3D and two groups of algorithms: classical algorithms (APF, RRT, and A*) and Q-learning-based algorithms (such as DFQL, IQL, and MEAEO-RL). The finding demonstrated the effectiveness of the proposed algorithm in terms of path length and computation time.

Neural network

Shiri et al.¹⁶⁵ proposed a neural network-based opportunistic Hamilton-Jacobi-Bellman (oHJB) approach for solving the remote UAV online path planning problem. The performance of oHJB was assessed using one base station and one UAV flying at a fixed altitude under wind constraints. Experimental results showed that oHJB gives better performance compared to other neural network-based algorithms in terms of path length, traveling time, and energy consumption.

Chhikara et al.¹⁶⁶ proposed a deep convolutional neural network (DCNN). Furthermore, genetic algorithms are utilized in CNN hyperparameter tuning. Experimental results demonstrated that the suggested algorithm gives better performance compared to the 11 state-of-the-art peer-to-peer deep neural networks.

Akshya et al.¹⁶⁷ developed a hybrid approach by combining deep neural networks (DNNs) with the Adam optimization algorithm for optimizing the UAV path planning. The performance of the proposed approach was evaluated by conducting experiments using different activation functions, including tanh, sigmoid, and ReLU. In the following, each generated trajectory is evaluated based on key metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R2). The results showed that the Adam-optimized DNN model outperformed the other activation functions in terms of path accuracy and smoothness.

Deep reinforcement learning (DRL)

Li and Wu.¹⁶⁸ proposed an enhanced DDPG algorithm for object avoidance and target tracking using an actor-recurrent-critic (ARC) architecture. The state of environments is approximated using long short-term memory networks. A predefined reward function based on the line of sight and an artificial potential field is designed so that the UAV can effectively maintain target tracking and avoid obstacles.

Zhang et al.¹⁶⁹ suggested a deep reinforcement learning algorithm based on LSTM networks to solve the problem of real-time path planning for UAVs. LSTM networks are used in this algorithm as Q-value networks for the deep Q network (DQN) algorithm, which makes the decision of the Q-value network has some memory. The simulation results showed that compared to traditional feed-forward neural networks (FNN), the RPP- LSTM proposed in this paper is significantly more accurate and robust than other real-time path planning algorithms.

Authors in this reference¹⁷⁰ proposed a deep reinforcement learning approach called Deep-State action reward state action (Deep-Sarsa) algorithm, which integrates classical Sarsa and neural networks to solve the problem of multi-UAV path planning. The performance of Deep-Sarsa technique was tested in a ROS-Gazebo simulation platform. The results demonstrated that Deep-Sarsa algorithm can offer a dependable path for UAVs, ensuring collision-free navigation.

Xi and Liu.¹⁷¹ developed a novel reinforcement learning (RL) training scheme based on DDPG for optimizing the UAV path planning. The testing of the proposed method, conducted in three phases, aimed at ensuring the UAV could reach its targets, avoid obstacles, and improve flight efficiency. Experimental results demonstrated the effectiveness of the proposed method compared to the synchronous RL training scheme and RRT* algorithm.

Yan et al.¹⁷² proposed an enhanced dueling double deep Q-networks (D3QN) algorithm. The performance of the improved (D3QN) technique was evaluated in dynamic and static environments. Experimental results proved that the D3QN algorithm outperformed DDQN and DQN in terms of path safety, reward accumulation, and success rate.

Critical analysis

Different machine learning-based UAV path planning algorithms were presented to show that these methods are able to manage complex dynamic environments. But still, these algorithms face several challenges, like high computational costs and long training time, which could become a barrier to convergence and stability. Table 6 provides a summary of machine learning (ML) techniques used in UAV path planning, detailing their applications along with their advantages and limitations.

Table 6.

Advantages and disadvantages of machine learning algorithms.

Algorithmtype	Advantages	Disadvantages	Applications
Supervised learning¹⁷³	- Appropriate for handling complex environments- Predicts with high accuracy	- High computational time requirements- Limited adaptability	- Complex and dynamic environments
Unsupervised learning^174,175	- Provides flexibility	- Less efficient in dynamic and real-time environments- It requires significant time	- Object detection- Environment clustering
Reinforcementlearning^164,176,177	- Strong adaptability in dynamic and uncertain environment	- High computational costs- Long training time	- Obstacle avoidance- Adaptive path optimization- Real-time path planning
Neural network^178–180	- Fast convergence speed- Effective in solving multi-objective UAV path planning	- Solutions are frequently suboptimal- High data requirements for effective training	- Object detection and classification- Obstacle avoidance
DRL^181,182	- Strong adaptability- Can effectively handle high-dimensional state representation	- High computational costs- Long training time	- Multi-objective optimization- Large-scale, unknown and complex environment

The limitations of ML algorithms have motivated researchers to propose a variety of methods, including hybrid algorithms,^183,184 adaptive strategies,^185,186 and novel improved ML algorithms.^187,188 Generally, three forms of hybridization in machine learning for UAV path planning can be identified: combining traditional algorithms with machine learning, for example, a combination of RRT with DQN.⁵⁸ RRT is popular due to its simplicity, robustness, and low computational time, but the limitation of the RRT method includes being a random sampling based. DQN is computationally expensive, difficult to tune, and not robust but offers adaptability in complex and dynamic environments. This integration enables the researchers to exploit advantages of both methods and eliminate the limitations of each technique. The second approach is integrating various machine learning algorithms; for instance, PPO is integrated with CNN and LSTM networks¹⁸⁹ to improve the convergence speed and increase the efficiency of avoiding collisions in complex environments. Finally, machine learning techniques are combined with computational algorithms, such as using an enhanced Gray Wolf Optimizer (GWO) for a viable solution for multi-UAVs in 3D environments, combined with reinforcement learning to guide the exploration and exploitation processes and enhance convergence speed.¹⁹⁰ Researchers also investigated the use of adaptive methods; Shi et al.¹⁹¹ introduced an Adaptive Dimensionality Reduction (ADR) framework, which incorporates Autoencoders (AEs) and Principal Component Analysis (PCA) to improve data processing efficiency and accuracy in Multi-Agent Reinforcement Learning (MARL). Finally, enhanced machine learning algorithms are also being created to overcome the limitations of standard ML methods. For example, Gong et al.¹⁹² used the Bayesian optimization method to enhance the learning efficiency, convergence speed, and policy stability of the traditional multi-agent deep deterministic policy gradient (MADDPG) algorithm for multi-UAV path planning. Wang et al.¹⁹³ proposed a Double Deep Q-Network (DDQN)-based algorithm for optimizing UAV path planning and conducted a comparative analysis with several reinforcement learning algorithms, including PPO, DQN, Q-learning, and Dueling DQN. The simulation results proved that DDQN and Dueling DQN exhibited better convergence performance and more stable learning behavior compared to PPO and DQN. The traditional DQN algorithm suffered from training instability due to overestimation. In contrast, PPO performed better than Q-learning but was limited by its lack of experience replay and high sensitivity to hyperparameters. Q-learning was the worst-performing algorithm. This is due to the fact that Q-learning requires storing every state-action pair’s value function, which can be enormously difficult in high-dimensional state space. Table 7 shows a collection of research studies that address problems in machine learning algorithms.

Table 7.

Recent improvements in machine learning algorithms for UAV path planning.

Algorithms	Limitations	Improved/hybrid algorithms	Solution methods
Supervisedlearning	- Slow training and limited adaptability	SLHS-SVM-AdaBoost¹⁹⁴	- Introduces an improved classification prediction model by combining SLHS with SVM-AdaBoost to improve the training efficiency and provide adaptability
Unsupervised learning	- Less efficient in dynamic and real-time environments- It requires significant time	Adaptive clustering¹⁹⁵	- Combines adaptive clustering with symbiotic organisms search optimization to improve adaptability in complex and environment and minimize the searching time
Reinforcementlearning	- High computational costs- Long training time	ACSQL¹⁹⁶	- Employs an adaptive conversion speed Q-learning algorithm by initializing the Q-table based on UAV sensor data to improve the training convergence
Reinforcementlearning	- High computational costs- Long training time	IR-QLA¹⁹⁷	- Introduces an instructed reinforcement Q-learning algorithm which uses received signal strength(RSS) as a dynamic reward metric to improve the learning speed and minimize the iteration counts needed for path optimization
Neural network	- Solutions are frequently suboptimal- High data requirements for effective training	IFDS-ANN¹⁹⁸	- Combines ANN with improved interfered dynamical system (IFDS) to obtain a sufficient amount of sample data for ANN training, enhancing path quality and computational efficiency
DRL	- Long training time- Slow learning in dynamic environment	Improved DQN¹⁹⁹	- Integrates deep Q-network with lazy training technique to reduce the training time
DRL	- Long training time- Slow learning in dynamic environment	Improved DRL²⁰⁰	- Introduces two learning stages: a reinforcement stage using DQN trained with the Bellman equation and a self-supervised stage for fine tune DQN’s backbone layers

Hybrid algorithms

Due to the diverse strengths and weaknesses of different path planning algorithms, researchers frequently investigate the integration of various algorithms in order to address their individual shortcomings and exploit their advantages, leading to the creation of hybrid algorithms. Using hybrid approaches, UAV route planning challenges can be met with greater robustness and effectiveness by combining different algorithms’ strengths. Typically, several methods of hybridization exist. This paper categorizes hybridization methods into three types: Traditional with Computational intelligence (TCi), Traditional with Machine learning (TMl), and Computational intelligence with Machine learning algorithms (CiM), as summarized in Table 8.

Table 8.

Summary of hybrid algorithms for UAV path planning.

Classification	Algorithm	Year	Env	Benchmark type	Single UAV	Results or performances
Traditional withcomputationalintelligence (TCI)	PSO with APF²⁰¹	2019	Dynamic	Simulated	No	Effectiveness of PSO-APF compared to PSO in terms of path length, path cost, and execution time.
	HGA²⁰²	2016	Static	Simulated	Yes	Robustness of HGA compared to CSA and CPLEX in terms of path length.
	PRM with ABC²⁰³	2021	Dynamic	Simulated	No	The efficiency of HPP compared to the PRM and traditional ABC schemes in terms of path length, energy consumption, flight time, and convergence time.
	MACO-MEA*²⁰⁴	2023	Dynamic	Simulated,Real-world	Yes	MACO-MEA* outperforms MACO-A* in terms of path length, execution time, and energy consumption.
Traditional with machinelearning (TML)	A* with Q-learning²⁰⁵	2022	Dynamic	Simulated	Yes	The effectiveness of the proposed algorithm in solving UAV path planning problems
	ANN with APF²⁰⁶	2023	Dynamic	Simulated	No	The efficiency of ANN-APF compared to the state-of-the-art methods.
	B-APFDQN²⁰⁷	2023	Dynamic	Simulated	Yes	The superiority of B-APFDQN compared to classical DQN in terms of path length and convergence time.
	Voronoi diagramwith Q-learning²⁰⁸	2024	Dynamic	Simulated	No	The effectiveness of the proposed algorithm compared to the PSO algorithm and the original Q-learning in terms of path length efficiency and collision avoidance.
Computational intelligence with machine learning (CiM)	RLGWO²⁰⁹	2020	Dynamic	Simulated	No	Effectiveness of RLGWO in a complex environment
	GMOPSO-QL²¹⁰	2021	Dynamic	Simulated	Yes	Robustness of GMOPSO-QL compared to MOPSO and GMOPSO algorithms
	SA-MAXQ²¹¹	2022	Dynamic	Simulated	No	The efficiency of SA-MAXQ compared to standard MAXQ, Q-Learning, and ε-Q-Learning algorithms in terms of convergence, time consumption, and search steps.

Traditional with computational intelligence (TCi)

Girija and Joshi²⁰¹ developed a novel technique by integrating particle swarm optimization (PSO) with APF algorithm to solve the problem of multi-UAVs path planning. The performance of the proposed algorithm was tested in a 3D dynamic environment with static and dynamic obstacles and a multi-UAV system. Simulation results demonstrated that PSO-APF algorithm generates collision-free paths, even in the presence of multiple obstacles, and performs better than the conventional PSO algorithm in terms of speed and path cost.

Arantes et al.²⁰² developed a hybrid technique, called Hybrid Genetic Algorithm (HGA), which is based on the combination of GA and Visibility Graph for solving the UAV path planning problem. The effectiveness of HGA was evaluated in a 2D non-convex environment using static obstacles distributed in 50 maps. Simulation results proved the effectiveness of HGA compared to CSA²¹² and CPLEX²¹³ in terms of path length efficiency and running time optimization.

Poudel and Moh²⁰³ suggested a hybrid path planning (HPP) technique by combining the probabilistic roadmap (PRM) algorithm and optimized artificial bee colony (ABC) algorithm to ensure effective data collection and avoid collisions for UAVs in emergency situations. In this method, the Probabilistic Roadmap Method (PRM) is employed to create the shortest and collision-free path, while an optimized artificial bee colony (ABC) technique is utilized to address various path constraints in a 3D environment. Based on simulation results, the proposed HPP performs noticeably better than the PRM and traditional ABC schemes in terms of path length, energy consumption, flight time, and convergence time.

Balasubramanian et al.²⁰⁴ proposed a hybrid approach, called modified ant colony optimization-MEA*, based on the combination of MACO and memory-efficient A* algorithm (MEA*) to find an energy-efficient optimal path while effectively avoiding obstacles in a 3D environment. Experimental results showed that the proposed algorithm outperforms MACO-A* in terms of path length, execution time, and energy consumption.

Traditional with machine learning approaches (TM)

Li et al.²⁰⁵ suggested a global-local hybrid algorithm that integrates the A* algorithm and Q-learning for optimizing the UAV path planning. In this method, the improved A* algorithm is used for global static planning, while the modified Q-learning technique is employed for local dynamic path adjustment for UAVs. The simulation results demonstrated that the proposed algorithm effectively addresses the challenges of UAV path planning.

Thangaraj and Sankar Sangam²⁰⁶ proposed a hybrid approach using artificial neural network for solving the UAV path planning problem, while obstacles are avoided by using artificial potential field (APF). In comparison to the state-of-the-art methods. Simulation results showed that the proposed method performs more effectively.

Kong et al.²⁰⁷ proposed a hybrid path planning algorithm (B-APFDQN) based on the hybridization of optimized Deep Q-Network (DQN) with Artificial Potential Field (APF) to address the UAV path planning problem. Experimental results showed that the proposed (B-APFDQN) algorithm outperforms the classical DQN in terms of path length and convergence time.

Su et al.²⁰⁸ hybridized Voronoi Diagrams and Q-learning algorithms for solving the multi-UAVs path planning problem in complex environments, with Voronoi Diagrams used to build a model that incorporates various obstacles within the UAVs operational area, and an enhanced Q-learning algorithm to improve path decision-making. Experimental results showed the performance of the suggested technique compared to the PSO algorithm and the original Q-learning in terms of path length efficiency and collision avoidance.

Computational intelligence with machine learning approaches (CiM)

Qu et al.²⁰⁹ hybridized GWO and reinforcement learning model for solving the UAV path planning problem. The performance of RLGWO was simulated in three case studies inside a 3D area with the presence of eight static obstacles. The experimental results demonstrated the effectiveness of RLGWO in a complex environment.

Xia and Zhang.²¹⁰ proposed a hybrid approach, called Gaussian multi-objective particle swarm optimization-QL algorithm (GMOPSO-QL), based on the combination of PSO with Q-learning algorithm for solving the constrained UAV path planning problem. In this paper, the Gaussian distribution is also introduced to generate new particles, and the effectiveness of the GMOPSO-QL algorithm was evaluated in a 3D environment. The simulation results showed that GMOPSO-QL outperforms both MOPSO and GMOPSO, demonstrating its effectiveness and robustness in achieving a safe and smooth path for UAVs.

Yuting et al.²¹¹ developed a hybrid approach, called SA-MAXQ based on the combination of MAXQ hierarchical reinforcement learning method with Simulated Annealing (SA) for solving the multi-UAV path planning problem. Experimental results proved the superiority of SA-MAXQ compared to standard MAXQ, Q-learning algorithm, and ε-Q-learning algorithm in terms of convergence, time consumption, and search steps.

Ablation discussion on hybrid algorithms

In order to understand the role of individual components in hybrid algorithms more clearly, an ablation analysis is valuable and was conducted by comparing hybrid models to the original single-method algorithms. For instance, the hybrid approach (PSO+APF), presented by Girija and Joshi²⁰¹ performed better than individual PSO, particularly in scenarios with high obstacle density and moving obstacles. When tested individually, the original PSO demonstrated the advantage of strong global search capability and reduced computational steps but struggled in complicated spaces. On the other hand, APF achieves fast local obstacle avoidance through modeling attractive and repulsive forces but is prone to getting stuck in local minima. Their integration successfully enhances the deficiencies of the two constituents: PSO’s global optimization defeats APF’s local traps, while APF enhances PSO’s convergence speed. Similarly, in the hybrid method suggested by Li et al.,²⁰⁵ an improved A* algorithm was combined with Q-learning. In this approach, global path planning was solved using the modified A* algorithm and the local dynamic path planning was handled by the modified Q-learning algorithm. Although A* has the advantage of being capable of planning globally optimal trajectories efficiently in static environments, it lacks the flexibility to adapt to dynamic or unexpected changes during the flight. Conversely, Q-learning is capable of dealing with local adaptations, but is plagued by slow convergence and inefficiency in large or complicated state spaces. By combining these two techniques, the authors realized a balanced solution that minimizes planning time, path complexity, and cost, while guaranteeing stable and efficient UAV path planning. As a third example from the CiM category, Qu et al.²⁰⁹ proposed the Reinforcement Learning-based Gray Wolf Optimizer (RLGWO), which integrates the capabilities of RL and the GWO to enhance four search strategies: exploration, exploitation, geometric adjustment, and optimal adjustment. Ablation results showed that compared to the original GWO and its improved variants (IGWO, MGWO, and EEGWO), the RLGWO algorithm outperformed them in generating viable and efficient routes, particularly in complicated environments.

Conclusion

Path planning stands as a pivotal challenge in deploying UAVs across diverse domains. In order to tackle this problem, this article provides a comprehensive overview of intelligent algorithms for UAV path planning, as proposed by a number of researchers in recent years through publications in major journals, and conference proceedings. It starts with a review of objectives and constraints of UAV path planning and then a brief discussion of classical approaches, followed by an analysis of the limitations and complexity of traditional methods that justify the shift toward intelligent techniques. These are divided into three classes: computational intelligence, machine learning techniques, and hybrid approaches. Each category has been divided into several subcategories. Further analysis of these categories reveals the pros and cons of each algorithm type, along with its applicability to diverse UAV path-planning scenarios. This paper has also pointed out recent developments in overcoming the limitations of these algorithms and has thus provided a foundation for the development of robust and efficient solution techniques for UAV path planning. As a result, future research should focus on adaptive path planning strategies that can respond to real-time changes, uncertainties, and dynamic environments, particularly with increasing complexities in real-world scenarios. In such conditions, the choice of a powerful and adaptive algorithm is crucial for computing an optimal, collision-free route between source and destination to ensure safe flight and overall mission success.

Footnotes

Handling Editor: Divyam Semwal

ORCID iDs

Dradoum Alaa

Khelassi Abdelmadjid

Lachekhab Fadhila

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Vega

Ramírez

Saiz

, et al. Multi-temporal imaging using an unmanned aerial vehicle for monitoring a sunflower crop. Biosyst Eng 2015; 132: 19–27.

Meivel

Maheswari

Remote sensing analysis of agricultural drone. J Indian Soc Remote Sens 2021; 49: 689–701.

Petrovski

Radovanović

Application of detection reconnaissance technologies use by drones in collaboration with C4IRS for military interested. Contemp Maced Def 2021; 21: 117–126.

Petrovski

Radovanović

Behlić

Application of drones with artificial intelligence for military purposes. In: 10th International Scientific Conference on Defensive Technologies (OTEH 2022), Belgrade, Serbia, 2022, pp.92—100.

Peng

, et al. Rapid safety monitoring and analysis of foundation pit construction using unmanned aerial vehicle images. Autom Constr 2021; 128: 103706.

Khan

Gupta

SK.

Emerging UAV technology for disaster detection, mitigation, response, and preparedness. J Field Robot 2022; 39: 905–955.

Borghetti

Caballini

Carboni

, et al. The use of drones for last-mile delivery: a numerical case study in Milan, Italy. Sustainability 2022; 14: 1766.

Konert

Smereka

Szarpak

The use of drones in emergency medicine: practical and legal aspects. Emerg Med Int 2019; 2019: 1–5.

Muñoz

López

Quevedo

, et al. Multi UAV coverage path planning in urban environments. Sensors 2021; 21: 7365.

10.

Aggarwal

Kumar

Path planning techniques for unmanned aerial vehicles: a review, solutions, and challenges. Comput Commun 2020; 149: 270–299.

11.

Zhao

Zheng

Liu

Survey on computational-intelligence-based UAV path planning. Knowl Syst 2018; 158: 54–64.

12.

Debnath

Omar

Latip

NBA

, et al. Correction to: A review on energy efficient path planning algorithms for unmanned air vehicles. In: Alfred

Lim

Ibrahim

AAA

(eds) Computational science and technology. Springer, Singapore, 2019, p.C1.

13.

Cao

Path planning approaches for unmanned aerial vehicle. Highlights Sci Eng Technol 2023; 76: 146–152.

14.

Luo

Tian

Wang

Research on unmanned aerial vehicle path planning. Drones 2024; 8: 51.

15.

Ait Saadi

Soukane

Meraihi

, et al. UAV path planning using optimization approaches: a survey. Arch Comput Methods Eng 2022; 29: 4233–4284.

16.

Liang

Juntong

Xiao

, et al. A literature review of UAV 3D path planning. In: Proceeding of the 11th World Congress on Intelligent Control and Automation. Shenyang, China, 2014, pp.2376–2381. IEEE

17.

Zhang

Huo

Chen

, et al. Hybrid path planning of a quadrotor UAV based on Q-learning algorithm. In: 2018 37th Chinese Control Conference (CCC). Wuhan: IEEE, 2018, pp.5415–5419.

18.

Lin

Saripalli

Sampling-based path planning for UAV collision avoidance. IEEE Trans Intell Transp Syst 2017; 18: 3179–3192.

19.

Shi

Hou

, et al. 3D dubins net-recovery path planning for fixed wing UAV. In: 2018 Chinese control and decision conference (CCDC), Shenyang, 2018, pp.604–610. IEEE.

20.

Wang

, et al. UAV path planning for object observation with quality constraints: a dynamic programming approach. Epub ahead of print 2023. DOI: 10.48550/ARXIV.2312.04943.

21.

Darbari

Gupta

Verma

. Dynamic motion planning for aerial surveillance on a fixed-wing UAV. In: 2017 international conference on unmanned aircraft systems (ICUAS), Miami, FL, 2017, pp. 488–497. USA: IEEE.

22.

Huang

Shi

, et al. Time-efficient coverage path planning for energy-constrained UAV. Wirel Commun Mob Comput 2022; 2022: 1–15.

23.

Modares

Ghanei

Mastronarde

, et al. UB-ANC planner: Energy efficient coverage path planning with multiple drones. In: 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, 2017, pp.6182–6189. IEEE.

24.

Zhang

Liu

, et al. Toward energy-efficient UAV-Assisted wireless networks using an artificial intelligence approach. IEEE Wirel Commun 2022; 29: 77–83.

25.

Angley

Ristic

Moran

, et al. Search for targets in a risky environment using multi-objective optimisation. IET Radar Sonar Navig 2019; 13: 123–127.

26.

Gao

Liu

, et al. An efficient path planning method for the unmanned aerial vehicle in highway inspection scenarios. Electronics 2023; 12: 4200.

27.

Liu

Zheng

Cai

Adaptive path planning for unmanned aerial vehicles based on bi-level programming and variable planning time interval. Chin J Aeronaut 2013; 26: 646–660.

28.

Chen

Adaptive path planning method for UAVs in complex environments. Int J Appl Earth Obs Geoinf 2022; 115: 103133.

29.

Fransen

KJC

van Eekelen

JAWM

Pogromsky

, et al. A dynamic path planning approach for dense, large, grid-based automated guided vehicle systems. Comput Oper Res 2020; 123: 105046.

30.

Kanehara

Kagami

Kuffner

, et al. Path shortening and smoothing of grid-based path planning with consideration of obstacles. In: 2007 IEEE International Conference on Systems, Man and Cybernetics, Montreal, QC, 2007, pp.991–996. Canada: IEEE.

31.

Barraquand

Langlois

Latombe

J-C.

Numerical potential field techniques for robot path planning. IEEE Trans Syst Man Cybern 1992; 22: 224–241.

32.

Bounini

Gingras

Pollart

, et al. Modified artificial potential field method for online path planning applications. In: 2017 IEEE Intelligent Vehicles Symposium (IV), Los Angeles, CA, 2017, pp.180–185. USA: IEEE.

33.

Schmid

Pantic

Khanna

, et al. An efficient sampling-based method for online informative path planning in unknown environments. IEEE Robot Autom Lett 2020; 5: 1500–1507.

34.

Jaillet

Cortés

Siméon

Sampling-based path planning on configuration-space costmaps. IEEE Trans Robot 2010; 26: 635–646.

35.

Gao

Zheng

, et al. Fast Path Planning for firefighting UAV based on A-Star algorithm. J Phys Conf Ser 2021; 2029: 012103.

36.

Mandloi

Arya

Verma

AK.

Unmanned aerial vehicle path planning based on A* algorithm and its variants in 3d environment. Int J Syst Assur Eng Manag 2021; 12: 990–1000.

37.

Cai

Xing

, et al. Path planning for UAV tracking target based on improved A-star algorithm. In: 2019 1st international conference on industrial artificial intelligence (IAI), Shenyang, 2019, pp.1–6. China: IEEE.

38.

Mkiramweni

Yang

, et al. Game-theoretic approaches for wireless communications with unmanned aerial vehicles. IEEE Wirel Commun 2018; 25: 104–112.

39.

Zhao

. The comparison of four UAV path planning algorithms based on geometry search algorithm. In: 2017 9th international conference on intelligent human-machine systems and cybernetics (IHMSC), Hangzhou: IEEE, 2017, pp.33–36.

40.

Dhulkefl

Durdu

TerziOğlu

Dijkstra algorithm using UAV path planning. Konya J Eng Sci 2020; 8: 92–105.

41.

Naazare

Ramos

Wildt

, et al. Application of graph-based path planning for UAVs to avoid restricted areas. In: 2019 IEEE international symposium on safety, security, and rescue robotics (SSRR). Würzburg, 2019, pp.139–144. Germany: IEEE.

42.

Blasi

D’Amato

Mattei

, et al. UAV path planning in 3-D constrained environments based on layered essential visibility graphs. IEEE Trans Aerosp Electron Syst 2023; 59: 2359–2375.

43.

Chen

The path planning algorithm studying about UAV attacks multiple moving targets based on Voronoi diagram. Int J Control Autom 2016; 9: 281–292.

44.

Chen

Path planning and cooperative control for multiple UAVs based on consistency theory and Voronoi diagram. In: 2017 29th Chinese control and decision conference (CCDC), Chongqing, 2017, pp.881–886. China: IEEE.

45.

Larasati

Wijayanti

ER.

A comparison of ant colony optimization and depth first search for solving unmanned aerial vehicle – ground vehicle routing problem in Humanitarian Logistics. J Ind Eng Educ 2023; 1: 3118–3127.

46.

Palanisamy

Vijayanathan

. Cluster based multi agent system for breadth first search. In: 2020 20th international conference on advances in ICT for emerging regions (ICTer), Colombo, Sri Lanka: IEEE, 2020, pp. 54–58.

47.

Chen

An adaptive rapidly-exploring random tree. IEEE/CAA J Autom Sinica 2022; 9: 283–294.

48.

Cui

Yan

Mutual information-based Multi-AUV path planning for scalar field sampling using multidimensional RRT. IEEE Trans Syst Man Cybern Syst 2016; 46: 993–1004.

49.

Kavraki

Svestka

Latombe

J-C

, et al. Probabilistic roadmaps for path planning in high-dimensional configuration spaces. IEEE Trans Robot Autom 1996; 12: 566–580.

50.

Budiyanto

Cahyadi

Adji

, et al. UAV obstacle avoidance using potential field under dynamic environment. In: 2015 international conference on control, electronics, renewable energy and communications (ICCEREC), Bandung: IEEE, 2015, pp. 187–192.

51.

Jayaweera

Hanoun

A dynamic artificial potential field (D-APF) UAV path planning technique for following ground moving targets. IEEE Access 2020; 8: 192760–192776.

52.

Liao

Zhang

, et al. UAV path planning model based on R5DOS model improved A-Star algorithm. Appl Sci 2022; 12: 11338.

53.

Quan

Han

Zhou

, et al. Survey of UAV motion planning. IET Cyber-Syst and Robotics 2020; 2: 14–21.

54.

Ghambari

Golabi

Jourdan

, et al. UAV path planning techniques: a survey. RAIRO-Oper Res 2024; 58: 2951–2989.

55.

Karaman

Frazzoli

Sampling-based algorithms for optimal motion planning. Int J Rob Res 2011; 30: 846–894.

56.

Mohamed

Ziedan

Gaafar

Artificial potential field approaches for indoor mobile robot path planning: a review. Egypt Int J Eng Sci Technol 2023; 4: 89–98.

57.

Zhang

, et al. Algorithm for UAV path planning in high obstacle density environments: RFA-star. Front Plant Sci 2024; 15: 1391628.

58.

Castro

GGRD

Berger

Cantieri

, et al. Adaptive path planning for fusing rapidly exploring random trees and deep reinforcement learning in an agriculture dynamic environment UAVs. Agriculture 2023; 13: 354.

59.

Khan

Moorthy

Alzubaidi

, et al. Computational intelligence for solving complex optimization problems. E3S Web Conf 2023; 399: 04038.

60.

Suganthan

, et al. Performance assessment and exhaustive listing of 500+ nature-inspired metaheuristic algorithms. Swarm Evol Comput 2023; 77: 101248.

61.

Deng

Zhang

, et al. A hybrid differential evolution particle swarm optimization algorithm based on dynamic strategies. Sci Rep 2025; 15: 4518.

62.

Xiao

Tan

Wang

A simulated annealing algorithm and grid Map-Based UAV coverage path planning method for 3D reconstruction. Electronics 2021; 10: 853.

63.

Xiao

Integrating machine learning for optimal path planning. J Comput Technol Appl Math 2025; 2: 21–26.

64.

Wang

EI-Hajjar

Deep reinforcement learning assisted UAV path planning relying on cumulative reward mode and region segmentation. IEEE Open J Veh Technol 2024; 5: 737–751.

65.

Ibrahim

Mostafa

Jnadi

, et al. Comprehensive overview of reward engineering and shaping in advancing reinforcement learning applications. IEEE Access 2024; 12: 175473–175500.

66.

Mehmood

Ali

, et al. A novel hybrid genetic and A-star algorithm for UAV path optimization. In: 2024 IEEE 1st Karachi section humanitarian technology conference (KHI-HTC), Tandojam, Pakistan: IEEE, 2024, pp.1–5.

67.

Jin

, et al. A new hybrid reinforcement learning with artificial potential field method for UAV target search. Sensors 2025; 25: 2796.

68.

Han

Zhang

A continuous space path planning method for unmanned aerial vehicle based on particle swarm optimization-enhanced deep Q-network. Drones 2025; 9: 122.

69.

Causa

Fasano

Grassi

Multi-UAV path planning for autonomous missions in mixed GNSS coverage scenarios. Sensors 2018; 18: 4188.

70.

Zhang

Liu

Feng

Research on obstacle avoidance path planning of UAV in complex environments based on improved Bézier curve. Sci Rep 2023; 13: 16453.

71.

Massaoudi

Berguiga

Harchay

, et al. Spectral and energy efficiency trade-off in UAV-based olive irrigation systems. Appl Sci 2023; 13: 10739.

72.

Cui

, et al. A UAV path planning algorithm for bridge construction safety inspection in complex terrain. Sci Rep 2025; 15: 13564.

73.

Zheng

Liu

3D UAV trajectory planning with obstacle avoidance for UAV-Enabled time-constrained data collection systems. IEEE Trans Vehicular Technol 2025; 74: 1460–1474.

74.

Wang

Zhang

Wang

, et al. Dynamic trajectory planning for Multi-AAV multimission operations using a Hybrid Strategy. IEEE Trans Aerosp Electron Syst 2025; 61: 7369–7386.

75.

Azar

Vaidyanathan

. Computational intelligence applications in modeling and control. 2015. Epub ahead of print. DOI: 10.1007/978-3-319-11017-2

76.

Nayeem

Fan

, et al. A modified particle swarm optimization for autonomous UAV path planning in 3D environment. In: Bhuiyan

Rahman

Ali

(eds) Cyber security and computer science. Springer International Publishing, 2017, pp.180–191.

77.

Sonny

Yeduri

Cenkeramaddi

LR.

Autonomous UAV path planning using modified PSO for UAV-assisted wireless networks. IEEE Access 2023; 11: 70353–70367.

78.

Zhang

Path planning of multirotor UAV based on the improved ant colony algorithm. J Robot 2022; 2022: 1–9.

79.

Lin

, et al. Improved artificial bee colony algorithm based on multi-strategy synthesis for UAV path planning. IEEE Access 2022; 10: 119269–119282.

80.

Rao

Wang

Shao

A multi-strategy collaborative grey wolf optimization algorithm for UAV path planning. Electronics 2024; 13: 2532.

81.

Wang

Chen

. Path planning and obstacle avoidance of unmanned aerial vehicle based on improved genetic algorithms. In: Proceedings of the 33rd Chinese Control Conference. Nanjing: IEEE, 2020, pp. 8612–8616.

82.

Yuan

Liu

Lian

, et al. Global optimization of UAV area coverage path planning based on Good Point Set and genetic algorithm. Aerospace 2022; 9: 86.

83.

Yin

Wang

Xiang

, et al. Adaptive differential evolution for collaborative path planning of multiple unmanned aerial vehicles. In: 2024 36th Chinese control and decision conference (CCDC). Xi’an, China: IEEE, 1998, pp.1521–1526.

84.

Zeng

Tong

Xia

Multi-UAV cooperative coverage search for various regions based on differential evolution algorithm. Biomimetics 2024; 9: 384.

85.

Huo

Zhu

, et al. A novel simulated annealing based strategy for balanced UAV task assignment and path planning. Sensors 2020; 20: 4769.

86.

Tong

Wang

, et al. Optimal route planning for truck–drone delivery using variable neighborhood Tabu search algorithm. Appl Sci 2022; 12: 529.

87.

Kumar

Garg

Singh

, et al. MVO-based 2-D path planning scheme for providing quality of service in UAV Environment. IEEE Internet Things J 2018; 5: 1698–1707.

88.

Chakraborty

Kar

AK.

Swarm intelligence: a review of algorithms. In: Patnaik

Yang

Nakamatsu

(eds.) Nature-inspired computing and optimization. Springer International Publishing, 2010, pp.475–494.

89.

Chen

, et al. Optimal energy consumption path planning for unmanned aerial vehicles based on improved particle swarm optimization. Sustainability 2023; 15: 12101.

90.

Yang

Chen

Optimization of dynamic obstacle avoidance path of multirotor UAV based on ant colony algorithm. Wirel Commun Mob Comput 2022; 2022: 1–9.

91.

Tian

Zhang

Bai

, et al. Real-time dynamic track planning of multi-UAV formation based on improved artificial bee colony algorithm. In: 2018 37th Chinese control conference (CCC). Wuhan: IEEE, 1997, pp.10055–10060.

92.

, et al. Path planning of UAV for oilfield inspection based on improved grey wolf optimization algorithm. In: 2019 Chinese control and decision conference (CCDC). Nanchang, China: IEEE, 2015, pp.3666–3671.

93.

Clerc

Kennedy

The particle swarm - explosion, stability, and convergence in a multidimensional complex space. IEEE Trans Evol Comput 2002; 6: 58–73.

94.

Shi

Eberhart

A modified particle swarm optimizer. In: 1998 IEEE international conference on evolutionary computation proceedings. IEEE world congress on computational intelligence (Cat. No.98TH8360). Anchorage, AK, USA: IEEE, 2015, pp.69–73.

95.

Hao

Luo

Zhang

Application study of UAV path planning based on the balanced search factor artificial bee colony algorithm. J Phys Conf Ser 2021; 2083: 032064.

96.

Zhu

Kwong

Gbest-guided artificial bee colony algorithm for numerical function optimization. Appl Math Comput 2010; 217: 3166–3173.

97.

Mirjalili

Lewis

The whale optimization algorithm. Adv Eng Softw 2016; 95: 51–67.

98.

Storn

Price

Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces. J Glob Optim 1997; 11: 341–359.

99.

Mirjalili

The ant lion optimizer. Adv Eng Softw 2015; 83: 80–98.

100.

Mirjalili

SCA: A sine cosine algorithm for solving optimization problems. Knowl Syst 2016; 96: 120–133.

101.

Bagherian

Alos

3D UAV trajectory planning using evolutionary algorithms: a comparison study. Aeronaut J 2015; 119: 1271–1285.

102.

Liu

Research on path planning algorithm for multi-UAV maritime targets search based on genetic algorithm. In: 2020 IEEE international conference on information technology, big data and artificial intelligence (ICIBA), Chongqing, China: IEEE, 2024, pp.840–843.

103.

Liu

Yang

, et al. An improved differential evolution based artificial fish swarm algorithm and its application to AGV path planning problems. In: 2017 36th Chinese control conference (CCC), Dalian, China: IEEE, 2015, pp.2556–2561.

104.

Sallam

Elsayed

Chakrabortty

, et al. Improved multi-operator differential evolution algorithm for solving unconstrained problems. In: 2020 IEEE congress on evolutionary computation (CEC), Glasgow, UK: IEEE, 2018, pp.1–8.

105.

Tanabe

Fukunaga

. Success-history based parameter adaptation for differential evolution. In: 2013 IEEE Congress on Evolutionary Computation. Cancun, Mexico: IEEE, 2019, pp. 71–78.

106.

Behnck

Doering

Pereira

, et al. A modified simulated annealing algorithm for SUAVs path planning. IFAC-PapersOnLine 2015; 48: 63–68.

107.

Balan

Luo

. Optimal trajectory planning for multiple waypoint path planning using Tabu search. In: 2018 9th IEEE annual ubiquitous computing, electronics & mobile communication conference (UEMCON), New York City, NY, 2022, pp.497–501. USA: IEEE.

108.

Jalali

SMJ

Khosravi

Kebria

, et al. Autonomous robot navigation system using the evolutionary multi-verse optimizer algorithm. In: 2019 IEEE international conference on systems, man and cybernetics (SMC), Bari, Italy: IEEE, 2018, pp.1221–1226.

109.

Mirjalili

Lewis

Grey wolf optimizer. Adv Eng Softw 2014; 69: 46–61.

110.

Mirjalili

Dragonfly algorithm: a new meta-heuristic optimization technique for solving single-objective, discrete, and multi-objective problems. Neural Comput Appl 2016; 27: 1053–1073.

111.

Mirjalili

Moth-flame optimization algorithm: a novel nature-inspired heuristic paradigm. Knowl Syst 2015; 89: 228–249.

112.

Tang

Xiao

, et al. Improved PSO-based two-phase logistics UAV path planning under dynamic demand and wind conditions. Drones 2024; 8: 356.

113.

Huang

A novel three-dimensional path planning method for fixed-wing UAV using improved particle swarm optimization algorithm. Int J Aerosp Eng 2021; 2021: 1–19.

114.

Peng

Chen

, et al. Online route planning for UAV based on model predictive control and particle swarm optimization algorithm. In: Proceedings of the 10th world congress on intelligent control and automation. Beijing, China: IEEE, 2022, pp.397–401.

115.

Calik

SK.

UAV path planning with multiagent ant colony system approach. In: 2016 24th signal processing and communication application conference (SIU). Zonguldak, Turkey: IEEE, 2024, pp.1409–1412.

116.

Cekmez

Ozsiginan

Sahingoz

. A UAV path planning with parallel ACO algorithm on CUDA platform. In: 2014 international conference on unmanned aircraft systems (ICUAS), Orlando, FL: IEEE, 2023, pp.347–354.

117.

Konatowski

Application of the ACO algorithm for UAV path planning. Electrotech Rev 2019; 1: 117–121.

118.

Muntasha

Karna

Shin

. Performance analysis on artificial bee colony algorithm for path planning and collision avoidance in swarm unmanned aerial vehicle. In: 2021 international conference on artificial intelligence and mechatronics systems (AIMS), Bandung, Indonesia: IEEE, 2020, pp.1–6.

119.

Tan

Shi

Gao

, et al. Multi-UAV path planning based on IB-ABC with restricted planned arrival sequence. Robotica 2023; 41: 1244–1257.

120.

Wang

Zhao

Han

, et al. A grey wolf optimizer using Gaussian estimation of distribution and its application in the multi-UAV multi-target urban tracking problem. Appl Soft Comput 2019; 78: 240–260.

121.

Zhang

, et al. Path planning of UAV based on improved adaptive grey wolf optimization algorithm. IEEE Access 2021; 9: 89400–89411.

122.

Shui

, et al. An improved grey wolf optimizer with weighting functions and its application to unmanned aerial vehicles path planning. Comput Electr Eng 2023; 111: 108893.

123.

Tao

Zhong

Gao

, et al. A study on path planning of unmanned aerial vehicle based on improved genetic algorithm. In: 2016 8th international conference on intelligent human-machine systems and cybernetics (IHMSC), Hangzhou, China: IEEE, 2024, pp.392–395.

124.

Huang

junwei

Qiao

, et al. Unmanned aerial vehicle path planning based on improved genetic algorithm. J Comput Appl 2021; 41: 390.

125.

Zhang

Duan

An improved constrained differential evolution algorithm for unmanned aerial vehicle global route planning. Appl Soft Comput 2015; 26: 270–284.

126.

Kok

Rajendran

Differential-evolution control parameter optimization for unmanned aerial vehicle path planning. PLoS One 2016; 11: e0150558.

127.

Pan

J-S

Liu

Chu

S-C.

A hybrid differential evolution algorithm and its application in unmanned combat aerial vehicle path planning. IEEE Access 2020; 8: 17691–17712.

128.

Zhou

Duan

, et al. Chaotic differential evolution approach for 3D trajectory planning of unmanned aerial vehicle. In: 2013 10th IEEE international conference on control and automation (ICCA), Hangzhou, China: IEEE, 2019, pp. 368–372.

129.

Turker

Sahingoz

Yilmaz

. 2D path planning for UAVs in radar threatening environment using simulated annealing algorithm. In: 2015 international conference on unmanned aircraft systems (ICUAS), Denver, CO: IEEE, 2021, pp.56–61.

130.

Ait-Saadi

Meraihi

Soukane

, et al. A novel hybrid chaotic aquila optimization algorithm with simulated annealing for unmanned aerial vehicles path planning. Comput Electr Eng 2022; 104: 108461.

131.

Wangsheng

Chong

Ruhua

Application of simulated annealing particle swarm optimization in complex three-dimensional path planning. J Phys Conf Ser 2021; 1873: 012077.

132.

Jarray

Al-Dhaifallah

Rezk

, et al. Path planning of quadrotors in a dynamic environment using a multicriteria multi-verse optimizer. Comput Mater Contin 2021; 69: 2159–2180.

133.

Jain

Yadav

Prakash

, et al. MVO-based path planning scheme with coordination of UAVs in 3-D environment. J Comput Sci 2019; 37: 101016.

134.

Liang

Zhang

Bai

3D path planning based on MMVO. In: 2021 33rd Chinese control and decision conference (CCDC), Kunming, China: IEEE, 2016, pp. 7385–7391.

135.

Shi

Huang

Jiang

, et al. Path planning optimization of intelligent vehicle based on improved genetic and ant colony hybrid algorithm. Front Bioeng Biotechnol 2022; 10: 905983.

136.

Feng

Sun

Zhang

, et al. A UAV path planning method in three-dimensional space based on a hybrid gray wolf optimization algorithm. Electronics 2023; 13: 68.

137.

Gong

Chen

Niu

, et al. Grid-based coverage path planning with NFZ avoidance for UAV using parallel self-adaptive ant colony optimization algorithm in cloud IoT. J Cloud Comput 2022; 11: 29.

138.

Zhao

Deng

, et al. Path planning of unmanned vehicles based on adaptive particle swarm optimization algorithm. Comput Commun 2024; 216: 112–129.

139.

Liu

A novel path planning method for aerial UAV based on improved genetic algorithm. In: 2023 Third International Conference on Artificial Intelligence and Smart Energy (ICAIS), Coimbatore, India: IEEE, 2013, pp.1126–1130.

140.

Wei

Improved particle swarm optimization based on multi-strategy fusion for UAV path planning. Int J Intell Comput Cybern 2024; 17: 213–235.

141.

Haghighi

Sadati

Dehghan

SMM

, et al. Hybrid form of particle swarm optimization and genetic algorithm for optimal path planning in coverage mission by cooperated unmanned aerial vehicles. J Aerosp Technol Manag 2020; 12: e4320.

142.

Saeed

Ali

Abdelhaq

, et al. Energy efficient path planning scheme for unmanned aerial vehicle using hybrid generic algorithm-based Q-learning optimization. IEEE Access 2024; 12: 13400–13417.

143.

Cao

Wang

, et al. An adaptive population size based differential evolution by mining historical population similarity for path planning of unmanned aerial vehicles. Inf Sci 2024; 666: 120432.

144.

Guo

Chen

. A path planning method for unmanned aerial vehicles based on improved artificial bee colony algorithm. In: 2023 7th international conference on electrical, mechanical and computer engineering (ICEMCE). Xi’an, China: IEEE, 2019, pp.665–669.

145.

Zhang

Han

Chen

, et al. A multi-strategy improved differential evolution algorithm for UAV 3D trajectory planning in complex mountainous environments. Eng Appl Artif Intell 2023; 125: 106672.

146.

Chu

Yang

Chaos particle swarm optimization enhancement algorithm for UAV safe path planning. Appl Sci 2022; 12: 8977.

147.

Tian

Shi

MPSO: modified particle swarm optimization and its applications. Swarm Evol Comput 2018; 41: 49–68.

148.

, et al. Trajectory planning for UAV based on improved ACO algorithm. IEEE Access 2020; 8: 2995–3006.

149.

Wang

Kong

An enhanced artificial bee colony algorithm for constraint optimization. Eng Lett 2024; 32.

150.

Huang

Luo

, et al. An improved fast convergent artificial bee colony algorithm for unmanned aerial vehicle path planning in battlefield environment. In: 2020 IEEE 16th international conference on control & automation (ICCA), Singapore: IEEE, 2020, pp.360–365.

151.

Xiao-Ting

Hai-Bin

, et al. Flight path planning based on an improved genetic algorithm. In: 2013 third international conference on intelligent system design and engineering applications. China, Hong Kong: IEEE, 2013, pp.775–778.

152.

Niu

Jiang

, et al. ERRT-GA: expert genetic algorithm with rapidly exploring random tree initialization for multi-UAV path planning. Drones 2024; 8: 367.

153.

Fan

Yan

Self-adaptive differential evolution algorithm with discrete mutation control parameters. Expert Syst Appl 2015; 42: 1551–1572.

154.

Huang

Sheng

Fang

, et al. A simulated annealing-particle swarm optimization algorithm for UAV multi-target path planning. In: 2022 2nd international conference on consumer electronics and computer engineering (ICCECE), Guangzhou, China: IEEE, 2022, pp.906–910.

155.

Chen

Fan

, et al. Unmanned Aircraft Vehicle Path planning based on SVM algorithm. In: Sun

Liu

(eds.) Foundations and practical applications of cognitive systems and information processing. Springer Berlin Heidelberg, 2014, pp.705–714.

156.

Muñoz

López

Quevedo

, et al. Gaussian processes and fast marching square based informative path planning. Eng Appl Artif Intell 2023; 121: 106054.

157.

Alos

Dahrouj

Detecting contextual faults in unmanned aerial vehicles using dynamic linear regression and K-nearest neighbour classifier. Gyroscopy Navig 2020; 11: 94–104.

158.

Radmanesh

Kumar

French

, et al. Towards a PDE-based large-scale decentralized solution for path planning of UAVs in shared airspace. Aerosp Sci Technol 2020; 105: 105965.

159.

Pierre

Zakaria

Pal

. Self-organizing map approach to determining compromised solutions for multi-objective UAV path planning. In: 2012 12th international conference on control automation robotics & vision (ICARCV), Guangzhou, China: IEEE, 2012, pp.995–1000.

160.

Tartaglione

Ariola

Obstacle avoidance via landmark clustering in a path-planning algorithm. In: 2018 Annual American control conference (ACC), Milwaukee, WI: IEEE, 2018, pp.2776–2781.

161.

Chen

Dong

Shang

, et al. Multi-UAV autonomous path planning in reconnaissance missions considering incomplete information: a reinforcement learning method. Drones 2022; 7: 10.

162.

Sonny

Yeduri

Cenkeramaddi

LR.

Q-learning-based unmanned aerial vehicle path planning with dynamic obstacle avoidance. Appl Soft Comput 2023; 147: 110773.

163.

Yan

Xiang

. A path planning algorithm for UAV based on improved Q-learning. In: 2018 2nd international conference on robotics and automation sciences (ICRAS). Wuhan: IEEE, 2018, pp.1–5.

164.

Wang

Zhao

, et al. APPA-3D: an autonomous 3D path planning algorithm for UAVs in unknown complex environments. Sci Rep 2024; 14: 1231.

165.

Shiri

Park

Bennis

Remote UAV online path planning via neural network-based opportunistic control. IEEE Wirel Commun Lett 2020; 9: 861–865.

166.

Chhikara

Tekchandani

Kumar

, et al. DCNN-GA: a deep neural net architecture for navigation of UAV in indoor environment. IEEE Internet Things J 2021; 8: 4448–4460.

167.

Sureshkumar

, et al. Enhancing UAV path planning efficiency through Adam-optimized deep neural networks for area coverage missions. Procedia Comput Sci 2024; 235: 2–11.

168.

Path planning for UAV ground target tracking via deep reinforcement learning. IEEE Access 2020; 8: 29064–29074.

169.

Zhang

Guo

Zheng

, et al. Real-time UAV path planning based on LSTM network. J Syst Eng Electronics 2024; 35: 374–385.

170.

Path D-SBM-U. Planning and obstacle avoidance in a dynamic environment. In: Advances in swarm intelligence: 9th international conference, ICSI 2018, Shanghai, China, June 17-22, 2018, Proceedings, Part II. Shanghay, China, 2020, p.102.

171.

Liu

Unmanned aerial vehicle trajectory planning via staged reinforcement learning. In: 2020 international conference on unmanned aircraft systems (ICUAS), Athens, Greece: IEEE, 2020, pp.246–255.

172.

Yan

Xiang

Wang

Towards real-time path planning through deep reinforcement learning for a UAV in dynamic environments. J Intell Robot Syst 2020; 98: 297–309.

173.

Shi

Dong

Zhang

Path planning of unmanned aerial vehicle based on supervised learning. In: 2022 IEEE 8th international conference on computer and communications (ICCC), Chengdu, China: IEEE, 2022, pp.747–751.

174.

Choi

Jimenez

Mavris

DN.

Two-layer obstacle collision avoidance with machine learning for more energy-efficient unmanned aircraft trajectories. Robot Auton Syst 2017; 98: 158–173.

175.

Chen

Zhang

, et al. A clustering-based coverage path planning method for autonomous heterogeneous UAVs. IEEE Trans Intell Transp Syst 2022; 23: 25546–25556.

176.

G-T

Juang

J-G.

UAV path planning and obstacle avoidance based on reinforcement learning in 3D environments. Actuators 2023; 12: 57.

177.

Dai

, et al. A lightweight reinforcement-learning-based real-time path-planning method for unmanned aerial vehicles. IEEE Internet Things J 2024; 11: 21061–21071.

178.

Padhy

Verma

Ahmad

, et al. Deep neural network for autonomous UAV navigation in indoor corridor environments. Procedia Comput Sci 2018; 133: 643–650.

179.

Khan

Tufail

Khan

, et al. A novel framework for multiple ground target detection, recognition and inspection in precision agriculture applications using a UAV. Un Sys 2022; 10: 45–56.

180.

Sanna

Godio

Guglieri

Neural network based algorithm for multi-UAV coverage path planning. In: 2021 international conference on unmanned aircraft systems (ICUAS), Athens, Greece: IEEE, 2023, pp.1210–1217.

181.

Wang

Zhang

, et al. Autonomous navigation of UAV in large-scale unknown complex environment with deep reinforcement learning. In: 2017 IEEE global conference on signal and information processing (GlobalSIP), Montreal, QC: IEEE, 2023, pp.858–862.

182.

Guo

Jiang

, et al. UAV navigation in high dynamic environments: a deep reinforcement learning approach. Chin J Aeronaut 2021; 34: 479–489.

183.

Xing

Carlson

Yuan

Optimize path planning for UAV COVID-19 test kits delivery system by hybrid reinforcement learning. In: 2022 IEEE 12th annual computing and communication workshop and conference (CCWC), Las Vegas, NV, USA: IEEE, 2023, pp.177–183.

184.

Zhong

Zhao

Luo

, et al. Hybrid path planning and following of a quadrotor UAV based on deep reinforcement learning. In: 2024 36th Chinese control and decision conference (CCDC), Xi’an, China: IEEE, 2024, pp.1858–1863.

185.

Westheider

Rückin

Popović

Multi-UAV adaptive path planning using deep reinforcement learning. In: 2023 IEEE/RSJ international conference on intelligent robots and systems (IROS), Detroit, MI, 2022, pp.649–656. USA: IEEE.

186.

Chu

Zhang

, et al. DRAL: Deep reinforcement adaptive learning for multi-UAVs navigation in unknown indoor environment. In: IEEE Transactions on Intelligent Transportation Systems, 2021, vol. 23, pp.25546–25556. Epub ahead of print 2024. DOI: 10.48550/ARXIV.2409.03930

187.

Zhang

Ren

, et al. A multi-area task path-planning algorithm for agricultural drones based on improved double deep Q-learning net. Agriculture 2024; 14: 1294.

188.

Chen

, et al. Improve exploration in deep reinforcement learning for UAV path planning using state and action entropy. Meas Sci Technol 2024; 35: 056206.

189.

Liang

Liu

Multi-UAV autonomous collision avoidance based on PPO-GIC algorithm with CNN-LSTM fusion network. Neural Netw 2023; 162: 21–33.

190.

Kumar

Singh

Tiwari

Novel reinforcement learning guided enhanced variable weight grey wolf optimization (RLV-GWO) algorithm for multi-UAV path planning. Wirel Pers Commun 2023; 131: 2093–2123.

191.

Shi

Zhao

Chen

, et al. Enhancing unmanned aerial vehicle path planning in multi-agent reinforcement learning through adaptive dimensionality reduction. Drones 2024; 8: 521.

192.

Gong

Wang

, et al. Bayesian optimization enhanced deep reinforcement learning for trajectory planning and network formation in multi-UAV Networks. IEEE Trans Vehicular Technol 2023; 72: 10933–10948.

193.

Wang

Jiang

, et al. Trajectory planning for UAV-assisted data collection in IoT network: a double deep Q network approach. Electronics 2024; 13: 1592.

194.

Zeng

Xiang

X-J

Dou

Y-P

, et al. UAV data link anti-interference via SLHS-SVM-AdaBoost algorithm: classification prediction and route planning. J Electron Sci Technol 2024; 22: 100279.

195.

Chen

Zhang

, et al. An adaptive clustering-based algorithm for automatic path planning of heterogeneous UAVs. IEEE Trans Intell Transp Syst 2022; 23: 16842–16853.

196.

Sun

, et al. An adaptive conversion speed Q-learning algorithm for search and rescue UAV path planning in unknown environments. IEEE Trans Vehicular Technol 2023; 72: 15391–15404.

197.

Muzammul

Assam

Ghadi

, et al. IR-QLA: machine learning-based Q-learning algorithm optimization for UAVs faster trajectory planning by instructed- reinforcement learning. IEEE Access 2024; 12: 91300–91315.

198.

Niu

Yan

Wang

, et al. 3D real-time dynamic path planning for UAV based on improved interfered fluid dynamical system and artificial neural network. Adv Eng Inform 2024; 59: 102306.

199.

Shin

Kim

, et al. Effective lazy training method for deep q-network in obstacle avoidance and path planning. In: 2017 IEEE international conference on systems, man, and cybernetics (SMC), Banff, AB: IEEE, 2024, pp.1799–1804.

200.

Samma

El-Ferik

Autonomous UAV visual navigation using an improved deep reinforcement learning. IEEE Access 2024; 12: 79967–79977.

201.

Girija

Joshi

Fast hybrid PSO-APF algorithm for path planning in obstacle rich environment. IFAC-PapersOnLine 2019; 52: 25–30.

202.

Arantes

MDS

Arantes

JDS

Toledo

CFM

, et al. A hybrid multi-population genetic algorithm for UAV path planning. In: Proceedings of the genetic and evolutionary computation conference 2016. Denver, Colorado: ACM, 2024, pp. 853–860.

203.

Poudel

Moh

Hybrid path planning for efficient data collection in UAV-aided WSNs for emergency applications. Sensors 2021; 21: 2839.

204.

Balasubramanian

Elangovan

Tamilarasan

, et al. Optimal energy efficient path planning of UAV using hybrid maco-mea* algorithm: theoretical and experimental approach. J Ambient Intell Humaniz Comput 2022; 14: 1–21.

205.

Yin

Wong

, et al. Quality-oriented hybrid path planning based on A* and Q-learning for unmanned aerial vehicle. IEEE Access 2022; 10: 7664–7674.

206.

Thangaraj

Sankar Sangam

Intelligent UAV path planning framework using artificial neural network and artificial potential field. IJEECS 2023; 29: 1192.

207.

Kong

Wang

Gao

, et al. B-APFDQN: a UAV path planning algorithm based on deep Q-network and artificial potential field. IEEE Access 2023; 11: 44051–44064.

208.

Gao

, et al. Enhanced multi-UAV path planning in complex environments with Voronoi-based obstacle modelling and Q-learning. Int J Aerosp Eng 2024; 2024: 1–14.

209.

Gai

Zhong

, et al. A novel reinforcement learning based grey wolf optimizer algorithm for unmanned aerial vehicles (UAVs) path planning. Appl Soft Comput 2020; 89: 106099.

210.

Xia

Zhang

Constrained path planning for unmanned aerial vehicle in 3D terrain using modified multi-objective particle swarm optimization. Actuators 2021; 10: 255.

211.

Yuting

Dongcheng

W. Eric

, et al. Multi-UAV collaborative path planning using hierarchical reinforcement learning and simulated annealing. Int J Performability Eng 2022; 18: 463.

212.

Blackmore

Ono

Williams

BC.

Chance-constrained optimal path planning with obstacles. IEEE Trans Robot 2011; 27: 1080–1094.

213.

Bliek1ú

Bonami

Lodi

Solving mixed-integer quadratic programming problems with IBM-CPLEX: a progress report. In: 2014 Proceedings of the twenty-sixth RAMP symposium, Tokyo, Japan, 2014.