Learning-Based Over-the-Air Integrated Sensing, Communication and Computation in UAV Swarm-Enabled Intelligent Transportation Systems

The Over-the-air Integrated Sensing, Communication, and Computation (Air-ISCC), supported by Unmanned Aerial Vehicles (UAVs), is a key technology for future 6G wireless networks. Air-ISCC can facilitate the mutual gain of communication, sensing, and computation functions. Equipping UAVs with sensing and communication units and computation resources empowers them to sense network environments and incorporate sensing information to provide computation offloading and mobile computing services. To optimize sensing, communication, and computation performance jointly, we present a multi-objective optimization framework in this paper. This framework jointly optimizes time slot scheduling, power control, resource allocation, and service association to maximize the service success of Air-ISCC while minimizing the energy consumption of UAVs. We transform the Air-ISCC problem into a sequential decision-making problem and propose a Proximity policy optimization-Based Intelligent Air-ISCC algorithm (PBIA) based on deep reinforcement learning. Leveraging the parallelization capability of the PBIA algorithm, we further propose training intelligent agents based on parallel deep reinforcement learning to realize autonomous decision-making of UAV swarm. Experimental results show that PBIA can learn effective policies with high learning efficiency and stability. Compared to baselines, PBIA significantly enhances the service success rate from 16.32% to 61.44%.