Collision-Aware Clustering for Enhanced Cooperative Perception in V2V Systems

Intelligent Transportation Systems (ITS) rely on connected vehicles to overcome problems such as occlusions and potential accidents, due to non-line-of-sight (NLoS) and other perception challenges. These challenges are magnified when explored in conjunction with communication network limitations, such as limited coverage (e.g., base station limitations) or simple packet collisions. Regardless of the reason behind information loss, the successfully received information should be prioritized to allow successful cooperative perception and accident avoidance. We address these issues by proposing a clustering algorithm that considers information relevance to the receivers and requires no extra communication overhead or network infrastructure. Four different information scoring functions are explored to reorganize data based on its perception relevance in the different clusters, with collision awareness being the focal metric for cluster formation. Our proposed technique achieves the best reduction in the number of packets used compared to existing state-of-the-art: ETSI CPM rules, Look Ahead, and Redundancy Mitigation algorithms. Moreover, thanks to its packet prioritization and reordering, the proposed algorithm outperforms these approaches in terms of the number of packets successfully received by more than 25%. Additionally, it achieves 13.1% and 19.8% enhancement in newly perceived objects, compared to the CPM rules, for the urban and highway scenarios, respectively. Lastly, due to a 6X and 5X improvement in information quality, based on the developed information scoring functions, compared to the baselines for the urban and highway scenarios, respectively.