Towards Efficient Perception Data Sharing: A Semantic Approach to Collective Perception

As vehicular networks evolve, efficiently sharing perception data while minimizing communication overhead remains a critical challenge. Traditional Collective Perception Service (CPS) strategies transmit large volumes of data, leading to network congestion and increased latency. In this work, we introduce the Semantic Collective Perception Service (Semantic CPS), an approach that optimizes message transmission by prioritizing the most relevant information for connected vehicles. Leveraging a knowledge graph and relevance-based filtering, our method ensures that only the most valuable data is transmitted to potential receivers. We evaluate the performance of Semantic CPS through simulations, comparing it against a conventional CPS strategy. Our results demonstrate that Semantic CPS reduces data transmission by over 62 %, highlighting that the effectiveness of information is highly dependent on the driving context. These findings highlight the potential of goal-oriented semantic communication to improve the scalability, efficiency, and reliability of cooperative perception systems.