Autonomous driving technology aims to enhance road safety and efficiency by reducing the risks inherent in human-driven vehicles. However, advanced systems (L3-L5) still face substantial technical challenges. Current research predominantly focuses on three core modules: perception (encompassing multi-sensor fusion and deep learning-based point cloud processing), planning (employing reinforcement learning and meta-heuristic algorithms for path optimization), and control (utilizing Model Predictive Control (MPC) and fuzzy decision models). Despite these advancements, limitations such as insufficient subsystem integration, limited adaptability to dynamic environments, and the lack of an ethical decision-making framework hinder further progress. This paper proposes an optimized organizational algorithm to improve real-time collaboration among multiple modules. It verifies security and efficiency through simulation tests and investigates lightweight models and standardized testing frameworks. Moving forward, it is essential to advance multi-agent coordination, strengthen generalization capabilities in complex scenarios, and establish a robust ethical framework to promote the widespread adoption of autonomous driving technology.

autonomous driving, perception, planning, multi-sensor fusion, subsystem collaboration

Mathematical modelling in actual problems of science and technics