Energy-efficient multiuser and multitask computation offloading optimization method

For dynamic application scenarios of Mobile Edge Computing (MEC), an Energy-efficient Multiuser and Multitask Computation Offloading (EMMCO) optimization method is proposed. Under the consideration multiuser multitask computation offloading, first, EMMCO takes into account existence dependencies among different tasks within implementation, abstracts these as a Directed Acyclic Graph (DAG), models offloading problem Markov decision process. Subsequently, task embedding sequence in DAG fed to RNN encoder-decoder neural network with combination attention mechanism, long-term are successfully captured by this scheme. Finally, Improved Policy Loss Clip-based PPO2 (IPLC-PPO2) algorithm developed, trained developed algorithm. The loss function IPLC-PPO2 utilized preference for training process, parameters continuously updated select optimal scheduling decisions. Simulation results demonstrate that proposed can achieve lower latency, reduce energy consumption, obtain significant improvement Quality Service (QoS) than compared algorithms under situations mobile edge network.