Deep Reinforcement Learning Aided Task Partitioning and Computation Offloading in Mobile Edge Computing

With the wave of the Internet of Things (IoT), a vast number of IoT devices are connected to wireless networks. To better support the Quality of Service of IoT devices with constrained resources, mobile edge computing (MEC) provisions computing resources at the network edge to process their tasks in proximity. In this work, we investigate task partitioning and computation offloading in collaborative MEC. Specifically, we propose a novel Deep Reinforcement Learning called Deep Deterministic with Dirichlet Policy Gradient (D3PG), which builds on Deep Deterministic Policy Gradient to partition tasks and perform task offloading efficiently. The developed model can learn to optimize multiple objectives, including maximizing the number of tasks processed before their deadlines and minimizing the energy cost. Simulation results are provided and demonstrate that the proposed D3PG scheme outperforms existing approaches.