Computation Offloading and Resource Allocation in Multi-Constraint Edge Environments Using DRL

<title>Abstract</title> Through deploying computation and storage as ubiquitous resources at the network edge, Mobile Edge Computing (MEC) is a promising paradigm to obtain low-delay and high-reliability service experience. However, computation offloading and resource allocation in MEC environments are challenging due to dynamic system state and variable user demand. Existing solutions can not well adapt to such dynamic MEC environments with multiple constraints, because they depend on prior knowledge, which is difficult to obtain in highly dynamic and time-varying MEC systems. To solve this issue, we propose a novel Joint computation Offloading and resource Allocation empowered by deep Reinforcement Learning (JOA-RL) method. For multi-user sequential tasks, the JOA-RL can generate appropriate schemes according to current computational resources and network conditions, aiming to improve the success rate of task execution while reducing the task execution delay and energy consumption. Meanwhile, the JOA-RL introduces a pre-processing mechanism of task priority, which makes it able to make fast decisions to find appropriate computing modes for different tasks. Extensive experiments verify the effectiveness and superiority of the JOA-RL method. The results show that the JOA-RL achieves a higher success rate of task execution and a better balance between delay and energy consumption than other benchmark methods, and a higher success rate of task execution.