Temporal-Difference Learning Aided Task Offloading in Air-Ground Networks

The current exponential growth in the number of smart devices and the limited computing power of terminal devices make it difficult to solve some complex problems. This study proposes a deep reinforcement learning-based task offloading algorithm in which Temporal-Difference learning is aided to satisfy the demand for low latency and high mobility to improve the quality of service for users. Multi-UAVs are used as nodes for edge computing in an air-ground network, and a multi-intelligent deep deterministic policy gradient algorithm is used to optimize the trajectory of the UAVs as well as to reduce the delay of edge computing and the total energy consumption of the UAVs. Simulation results show that the energy consumption of the UAV is reduced by 14.65%, the average delay of task processing is reduced by 4.06%, and the rate of completing tasks increases by 3.16%-7.86%.