UAV-Aided Partial Task Offloading for Integrated Sensing, Computation, and Communications Systems via Deep Reinforcement Learning

The emergence of Metaverse applications has driven the integrated sensing, computation, and communications (ISC2) technology into edge networks to optimize performance. Unmanned aerial vehicles (UAVs) have become a feasible and cost-effective platform for improving ISC2 coverage and enabling low-latency applications due to high mobility. In this paper, we present a UAV-aided ISC2 system that employs partial data offloading to enhance the sensing data processing capabilities of edge network users. We propose a scheme that uses deep reinforcement learning to minimize the sum of communication and computation delays in the system. To this end, we optimize the data offloading and UAV's flight action based on deep deterministic policy gradient. Extensive simulations are conducted to evaluate the performance of our proposed method under various parameter conditions, which demonstrate its significant superiority over other methods.