Traffic-Burst-Resilient Queue-Adaptive Load Balancing for Edge-Assisted Autonomous Vehicles
Edge-assisted autonomous driving enables vehicles to offload computationally intensive perception tasks to vehicular edge computing (VEC) servers, thereby reducing on-board power consumption while maintaining real-time performance. However, in practical driving environments, burst traffic and shared workloads among multiple services can significantly increase queue backlogs, potentially degrading system stability and violating latency constraints. In this paper, we propose a traffic-burst-resilient queue-adaptive load balancing algorithm for edge-assisted autonomous vehicles. The proposed method jointly determines task offloading decisions and vehicle central processing unit (CPU) clock frequency using a Lyapunov optimization framework. To enhance robustness under timevarying traffic conditions, we introduce a dynamic trade-off parameter that adaptively adjusts the emphasis between energy efficiency and queue stability based on the current backlog state. When burst traffic causes rapid queue accumulation, the proposed scheme temporarily reduces the energy penalty weight to prioritize backlog stabilization. Simulation results demonstrate that the proposed dynamic parameter design maintains nearly the same level of power consumption as a fixed-parameter baseline, while reducing the average queue backlog by approximately 32%, thereby improving system stability under burst traffic conditions.