V2I Interference Management for Collaborative Perception Based on Hierarchical Reinforcement Learning

Collaborative perception technology contributes to the exchange of perception information between connected automated vehicles (CAVs) and roadside units (RSUs) to overcome the limitations of standalone perception and enhance the environmental perception accuracy of collaborators. This paper delves into the interference management issues in vehicular communications for collaborative perception and designs a hierarchical reinforcement learning model-driven resource allocation strategy. The uniqueness of this strategy lies in its comprehensive consideration of spatial confidence and channel state information (CSI) to optimize the efficiency of collaborative perception. Simulation experiments demonstrate that our proposed algorithm significantly improves perception accuracy compared to various baseline methods, especially under bandwidth constrained conditions.