Partial Computation Offloading and Resource Allocation via Deep Deterministic Policy Gradient

Abstract

Recent years have seen an explosive growth of mobile Internet applications, with a plethora of computation-intensive and latency-sensitive services running on the Internet of Things (IoT), posing a great challenge to its limited network resources. Computation offloading technology, as a hot direction in the field of mobile edge computing (MEC), can provide a reliable means to achieve efficient computation migration strate-gies. In this paper, we focus on optimizing the task offloading and resource allocation problem in the MEC system by a deep deterministic policy gradient (DDPG). For our simulated deployment of a single MEC server and multi-user scenario, we design a task cache queue for each terminal user and define the allocation ratio vectors of task offloading and resource allocation. By minimizing the weighted sum of the total time latency and the energy consumption, an optimal solution can be achieved via the DDPG. Experimental results show that the proposed scheme performs better in reducing total system overhead than the baselines.