A deep reinforcement learning resource allocation strategy for integrated sensing, communication and computing

Abstract

The growing number of sensor nodes requires higher computational and communication resources to meet the needs of more complex applications as increasingly sophisticated mobile communications and networks evolve. As energy and time are necessary to request these resources, Edge Computing(EC) has become a key enabler to solve energy and time consumption problems in the field of Integrated Sensing, Communication, and Computing (ISCC). Computation offloading is an important technology, and Edge devices (EDs) can be used to reduce latency and energy consumption by offloading tasks to EC servers. However, the transmission also causes latency and energy consumption, so this paper proposes a DDORS scheme to solve this problem by decomposing the original problem into two subproblems and using convex optimization and DRL to obtain optimal solutions to reduce the weighted total cost of completing tasks in terms of latency and energy consumption. The simulation results indicate that the proposed DDORS may drastically reduce costs when compared to various baseline schemes.