Age of information-aware multi-agent task scheduling strategy for vehicular edge computing

The rapid proliferation of intelligent transportation applications has created unprecedented demands for real-time data processing in vehicular networks. Vehicular Edge Computing (VEC) addresses these challenges by leveraging computing resources on moving vehicles, but ensuring information timeliness remains problematic due to network dynamics and resource constraints. This paper proposes a novel multi-agent scheduling framework that utilizes Age of Information (AoI) as the primary metric to maintain data freshness in VEC environments. Unlike conventional approaches that focus solely on latency reduction, our strategy employs a distributed multi-agent architecture where vehicle-mounted servers collaborate through limited information exchange to make coordinated scheduling decisions. We develop a hybrid optimization model combining game theory and reinforcement learning that enables agents to balance individual and collective objectives while adapting to vehicular mobility patterns. The scheduling mechanism operates across three dimensions: spatial (server selection), temporal (execution timing), and priority-based (task importance). Simulation results using real-world urban traffic datasets demonstrate that our approach reduces average AoI by 37 % compared to centralized scheduling methods, while improving computational resource utilization by 42 % and decreasing communication overhead by 29 %. The proposed framework shows particular effectiveness in dense urban scenarios and heterogeneous task environments, offering a scalable solution for maintaining information freshness in next-generation intelligent transportation systems.