Dynamic Computation Offloading with Deep Reinforcement Learning in Edge Network

With the booming proliferation of user requests in Internet Things (IoT) network, Edge Computing (EC) is emerging as a promising paradigm for provision flexible and reliable services. Considering resource constraints IoT devices, some delay-aware requests, heavy-workload device may not respond on time. EC has sparked popular wave offloading to edge servers at network. The orchestration user-requested schemes creates remarkable challenge regarding delay energy consumption devices networks. To solve this challenge, we propose dynamic computation strategy consisting following: (i) concept intermediate nodes, which can minimize current tasks handled by dynamically combining task-offloading service migration strategies; (ii) based workload node selection problem modeled multi-dimensional Markov Decision Process (MDP) space, deep reinforcement learning algorithm implemented reduce large MDP space make fast decision. Experimental results show that superior existing baseline methods delays devices.