Deep Reinforcement Learning-Based Task Unloading Algorithm in MEC

MEC is a new form that provides rich computing capability closing to the network edge to get closer to the users. In mobile edge computing, the mobile device can offload some tasks to the edge server for data storage and calculation. This paper assumes that there are a mobile device, multiple edge servers, and each edge server has limited resources. In order to improve the task offloading and assignment efficiency of MEC and decrease the equipment energy consumption and task execution delay, this paper proposes a task offloading algorithm based on modified deep reinforcement learning to select a best edge server for each offloading task. Firstly, the problem is described as a Markov decision process, a system model is established, and then deep reinforcement learning algorithms are used to solve the optimal task offloading strategy. The task offloading method based on deep reinforcement learning improves the algorithm convergence speed by setting the experience relay memory to quickly update the parameters in the network. Experimental results show that the algorithm improves the efficiency of task offloading compared with the commonly used basic algorithms, reduces the energy consumption of equipment and the time delay about task execution, and proves the reliability of the algorithm.