REAL-TIME PERFORMANCE OPTIMIZATION FOR AUTONOMOUS DRIVING SYSTEMS BASED ON EDGE COMPUTING ARCHITECTURE

Abstract

Traditional cloud-centric architectures often suffer from high latency, limiting their effectiveness in autonomous driving applications. This study introduces an edge computing-based optimization framework that enhances real-time responsiveness through a hierarchical task offloading strategy across collaborative edge nodes. Perception and decision-making modules are modularized using Docker containers to ensure lightweight encapsulation, while Kubernetes is adopted for dynamic resource scheduling and scalable deployment. The proposed system is validated on the Baidu Apollo autonomous driving platform. Experimental results show a 23.6% reduction in end-to-end latency with only a 2.8% decrease in mean Average Precision (mAP) for object detection. The architecture also demonstrates strong scalability and deployment flexibility, offering practical value for engineering-level implementations of autonomous driving systems.