A DRL-based algorithm for Dependent Task Offloading in Multi-access Edge Computing

As a promising technology, Multi-access Edge Computing (MEC) provides various resource to the smart devices (SD) with limited ability for faster task completion. While independent tasks are widely investigated in the literature, in reality, the tasks are usually correlative in an application. In this paper, we investigate the offloading problem of dependent tasks in a varying wireless condition. Our object is to minimize the total completion time of all dependent tasks. To solve the problem, we formulate the dependent tasks as a Directed Acyclic Graph (DAG) and propose a deep reinforcement learning based method to make offloading decision quickly. We conduct several simulation experiments and the results show that our method achieves lower time consumption than other baseline algorithms.