Editorial

The importance of sensor networks and the integration of signal processing have increased as a consequence of the growth of complex Internet-of-Things (IoT), distributed and wireless network applications, for commercial, medical, context-aware, and industrial domains, among others. The complexity, heterogeneity, and dynamicity of some sensor networks demand new intelligent solutions for data aggregation, integration, and management. Artificial intelligence helps in these sensor network applications to respond to the above-mentioned challenges.

In consideration of these issues, this Special Collection offered a platform for researchers to publish recent and original works in different topics, focusing on intelligent systems for sensor networks. Several researchers from different parts of the world submitted their papers. Hence, after a rigorous review process, we accept only seven papers for this collection. An overview of the key contributions of each paper is presented as follows.

Medical and healthcare services based on IoT need a high degree of autonomy. Developing an efficient sensing method of physical body data is challenging given that they are highly diverse and dynamic, especially on the cloud computing system for artificial intelligence computation. The first paper titled “An edge cloud-based body data sensing architecture for artificial intelligence computation” presents an edge cloud–based body data sensing architecture with the purpose of analyzing physical body data on an edge cloud computing system. The authors, Kim and Lim, aim to identify relationships between the person’s activities and health conditions with a framework, which efficiently aggregates and processes sensor data. They address the important challenges of providing real-time service and mobility and evaluated the effectiveness of their architecture for activity recognition based on body sensor data.

In other domains like face recognition, there are also important challenges to enhance these systems in an uncontrolled environment. The second paper contributed by Yu et al. entitled “A novel framework for face recognition using robust local representation-based classification” addresses the problem of how to enhance the performance of sparse representation classification-based face recognition systems in an unconstrained environment. They adopted a three-dimensional (3D)-based frontalization on the aligned downsampling local binary pattern feature to deal with the uncontrolled environments effectively. An optimized projection/sensing matrix is also designed in order to reduce the complexity and prevent overfitting problem.

Wireless sensor networks (WSNs) may also change dynamically due to external or internal factors. Therefore, an energy-efficient data aggregation method is needed to enhance wireless sensor nodes’ lifetime and quality of service. With this regard, the third paper “A review on the applications of multiagent systems in wireless sensor networks,” Derakhshan and Yousefi review recent simulated approaches and real-time applications of multiagent systems (MASs) in WSN. First, they offer an overview on simulated and real-time approaches using MAS in WSN. The key challenges, efficiency factors, limitations, and future directions are provided. According to the authors, data aggregation is one of the most important challenges in applications using MAS in WSN. They propose a framework for energy-efficient and secure data aggregation of this kind of applications.

Yang et al. present the fourth paper “Research on shore-based intelligent vessel support system based on multi-source navigation sensors simulation.” In this work, the authors propose an efficient simulation model for multi-source navigation sensors. They developed a virtual intelligent vessel as platform of a shore-based support system for remote monitoring of the autonomous navigation system. The virtual intelligent vessel platform is developed and tested with the vessel “Chang Shan Hai” and proved to be effective.

In the fifth paper “On the adaptability of ensemble methods for distributed classification systems: a comparative analysis,” Villaverde et al. propose a two-stage classifier ensemble architecture composed of various classifiers in order to combine estimations from sensors to obtain more reliable solutions and adapt to changing environment. In the first stage, classifications are made using artificial neural networks from partial information, and in the second stage, all estimations are aggregated by an ensemble in order to obtain the final classification. The authors compared four different algorithms in two example applications: pedestrian and car detection application and multi-sensor Modified National Institute of Standards and Technology (MNIST) (database of handwritten digits) application. They prove the applicability of the classifier ensemble for sensor cooperation with the first example. They measure the adaptability of the proposed algorithms testing the second application against different perturbations. This work can be useful for sensor data fusion applications with changing environments.

An important issue on distributed sensor networks is the reconfiguration of nodes (intelligent sensors) while they are dynamically moving through the net. In the article titled “Generation of broadcasting for fractal adaptive Internet of things reconfiguration under the swarm intelligence paradigm” by Moreno et al., the authors propose an adaptive reconfiguration algorithm to link the nodes in the network to a fractal topology. This proposal improves the reconfiguration of nodes and the communication among them using only two adjacent nodes to pass the information to the entire network. A fractal Hilbert topology is implemented in combination with swarm intelligence–based methods to finally link the nodes. This work promotes energy and communication efficiency through the sensor network replacing the use of a central node by a distributed configuration.

Human activity recognition (HAR) tries to understand people’s actions in different applications dealing with the integration of sensors. In the final paper titled “A concise review on sensor signal acquisition and transformation applied to human activity recognition and human-robot interaction,” Martinez and Ponce present a review on different steps of signal processing in the context of HAR. The authors analyze two applications in which HAR plays an important role to understand and implement computer interactions with humans in robotics: human–robot interaction (HRI) and socialization, and imitation learning. Ideas and trends regarding important design issues in the context of HAR for HRI are discussed. Although the high-quality review process restricted the number of accepted papers to seven, we hope that researchers in the community of sensor network can benefit from the authors’ innovative intelligent solutions, analysis, and future trends presented in this collection.