Collaborative Collision Avoidance for CAVs in Unpredictable Scenarios

Abstract

Modern connected and automated vehicles (CAV) are capable of making informed decisions in unexpected situations. CAVs can achieve this by collaborating with other CAVs using communication and sensing capabilities. This work discusses a partially-decentralized collaborative decision making approach for a coalition of CAVs in the presence of a misbehaving vehicle. A novel algorithm based on Monte Carlo Tree Search (MCTS) is presented for the CAV’s planning problem of deriving mitigation action plans. This algorithm reduces the size of the search tree exponentially to overcome the computational limitations of MCTS for large action-agent sets. V2V communication is used to ensure that mitigation action plans chosen by coalition members are conflict-free when possible. The proposed method is evaluated for several conflict scenarios showing that the system can effectively avoid collisions in diverse situations.