Integrated framework for path re-planning and tracking based on fusion prediction for autonomous obstacle avoidance

Abstract

Path planning and tracking are critical for ensuring the safety and efficiency of autonomous vehicles. Environmental uncertainties and path tracking errors significantly affect these processes. Traditional methods often struggle to effectively handle dynamic obstacles and ensure precise tracking in uncertain environments. To address these challenges, this paper introduces a novel framework that integrates path re-planning with path tracking control, combining both model-based and data-driven prediction methods. By using $H_{\infty }$ filter for dynamic obstacle prediction, this approach significantly enhances both predication accuracy and robustness compared to traditional methods. Based on these predictions, a local obstacle-avoidance path is generated using a constrained point mass model and fifth-degree polynomials. Furthermore, Model Predictive Control (MPC) is applied for high-precision path tracking. Simulations and Hardware-in-the-Loop (HiL) tests conducted in scenarios involving moving obstacles demonstrate that this framework effectively addresses emergency obstacle avoidance path planning while ensuring high-precision tracking.