Abstract:Cooperative perception (CP) is a key technology to facilitate consistent and accurate situational awareness for connected and autonomous vehicles (CAVs). To tackle the network resource inefficiency issue in traditional broadcast-based CP, unicast-based CP has been proposed to associate CAV pairs for cooperative perception via vehicle-to-vehicle transmission. In this paper, we investigate unicast-based CP among CAV pairs. With the consideration of dynamic perception workloads and channel conditions due to vehicle mobility and dynamic radio resource availability, we propose an adaptive cooperative perception scheme for CAV pairs in a mixed-traffic autonomous driving scenario with both CAVs and human-driven vehicles. We aim to determine when to switch between cooperative perception and stand-alone perception for each CAV pair, and allocate communication and computing resources to cooperative CAV pairs for maximizing the computing efficiency gain under perception task delay requirements. A model-assisted multi-agent reinforcement learning (MARL) solution is developed, which integrates MARL for an adaptive CAV cooperation decision and an optimization model for communication and computing resource allocation. Simulation results demonstrate the effectiveness of the proposed scheme in achieving high computing efficiency gain, as compared with benchmark schemes.

What problem does this paper attempt to address?

This paper primarily focuses on the issue of Cooperative Perception (CP) for Connected and Autonomous Vehicles (CAVs) in mixed traffic scenarios, aiming to improve perception efficiency and accuracy. ### Research Background and Objectives With the advancement of sensing technology, Artificial Intelligence (AI), and Vehicle-to-Everything (V2X) communication technology, the development of autonomous vehicles is gradually transforming the future transportation system, enhancing road safety and traffic efficiency. To improve the reliability and accuracy of environmental perception, Cooperative Perception technology has been proposed to supplement Stand-Alone Perception (SP), which involves sharing perception information through vehicle-to-vehicle communication. ### Problems Addressed - **Network Resource Utilization Efficiency**: Traditional broadcast-based cooperative perception suffers from low network resource utilization efficiency, especially in dense traffic conditions. - **Dynamic Environment Adaptability**: Considering the mobility of vehicles and the variability of available radio resources, achieving efficient resource allocation and supporting cooperative perception for all selected CAV pairs in a dynamic environment is a challenge. ### Main Contributions - Proposed an adaptive cooperative perception scheme that allows each CAV pair to dynamically decide whether to engage in cooperative perception based on the current environment. - Defined a joint optimization problem that includes two sub-problems: cooperative decision-making and resource allocation, aiming to maximize computational efficiency gains while minimizing the cost of mode switching. - Developed a model-assisted Multi-Agent Reinforcement Learning (MARL) solution to learn optimal cooperative decisions, combined with a model-based approach for resource allocation. ### Methodology - The paper considers feature-level cooperative perception, where vehicles transmit compressed feature data instead of raw data to reduce communication overhead. - Proposed an adaptive cooperative perception scheme that includes selecting which CAV pairs to engage in cooperative perception and dynamically allocating communication and computational resources to these pairs. - To balance transmission and computation delays, a portion of the available radio resources is dynamically allocated to support feature data transmission, and the CPU frequency on the vehicles is adjusted as needed. - Utilized a Multi-Agent Reinforcement Learning approach to learn optimal cooperative decisions, while employing a model-based method to solve the resource allocation problem given the cooperative decisions. ### Conclusion By proposing an adaptive cooperative perception scheme and corresponding optimization algorithms, this paper effectively improves the resource utilization efficiency and computational efficiency gains of cooperative perception in dynamic environments. This has significant implications for the future development of autonomous vehicles.

Model-Assisted Learning for Adaptive Cooperative Perception of Connected Autonomous Vehicles

MACP: Efficient Model Adaptation for Cooperative Perception

Scalable and Dynamic Cooperative Perception: A Data/Model Co-Driven Framework

Decentralized Multiagent Reinforcement Learning-based Cooperative Perception with Dual-functional Radar-Communication V2V Links

Soft Actor–Critic-Based Multilevel Cooperative Perception for Connected Autonomous Vehicles

PACP: Priority-Aware Collaborative Perception for Connected and Autonomous Vehicles

Enhanced Cooperative Perception for Autonomous Vehicles Using Imperfect Communication

Cooperative Perception with Deep Reinforcement Learning for Connected Vehicles

MASS: Mobility-Aware Sensor Scheduling of Cooperative Perception for Connected Automated Driving

Communication-Efficient Decentralized Multi-Agent Reinforcement Learning for Cooperative Adaptive Cruise Control

Consensus-Based Distributed Cooperative Perception for Connected and Automated Vehicles

RSU-Aided Energy-Efficient Collaborative Perception for Connected Autonomous Vehicles

Meta Learning Based Adaptive Cooperative Perception in Nonstationary Vehicular Networks

COOPERNAUT: End-to-End Driving with Cooperative Perception for Networked Vehicles

Delay-Aware Cooperative Perception with Deep Reinforcement Learning in Vehicular Networks

Cooperative Intersection with Misperception in Partially Connected and Automated Traffic

InterCoop: Spatio-Temporal Interaction Aware Cooperative Perception for Networked Vehicles

Shared Information-Based Safe And Efficient Behavior Planning For Connected Autonomous Vehicles

Accuracy-Aware Cooperative Sensing and Computing for Connected Autonomous Vehicles

A Cooperative Perception Environment for Traffic Operations and Control