Research On Task Cooperative Offloading For Mobile Edge Computing

With the continuous progress of society and technology,Mobile Device(MD)is growing exponentially in number.Although mobile smart devices are constantly updated and iterated to make great progress,they still face shortage of computing power,short battery life and high energy consumption in the face of computation-intensive and latency-sensitive applications.Mobile Edge Computing(MEC)technology can reduce the computation time of tasks and reduce the energy consumption of Io T mobile devices by offloading the local computation tasks from Io T devices with low computing power to edge servers with high computing power to complete the tasks.In this paper,we study the cooperative offloading technique of MEC technology,using the neighboring devices around MDs as relays to assist in achieving computational task offloading,jointly utilizing the computational resources of local MDs,relay MDs and MEC servers to achieve the optimal overall computational energy consumption of computational tasks,and providing some empirical references for the cooperative offloading research of MEC systems.The paper first investigates a single relay cooperative offload scenario,where the scenario consists of a local MD,a set of neighboring MDs,and multiple wireless access points(AP),each of which deploys a MEC server to provide rich computational resources.One of the neighboring MDs can act as a Relay Node(RN)for offloading the computational tasks of the local MD through a Device-to-Device(D2D)communication link.Based on this scenario,an energy-minimizing cooperative offloading scheme under the delay constraint is proposed.The local MD can offload part of the task segment of a computational task to the collaborating RN or AP,and the RN can also offload part of the task segment to the AP.The energy-minimizing cooperative offloading problem under the computational delay constraint is solved by a convex optimization method,and the resource allocation scheme,the selection of RN and AP,and the task segment partition for cooperative offloading are determined.The simulation results show that the energy consumption of this scheme is reduced by 90.3% and the maximum supported task length is increased by 83% compared to the local computation-only scheme.Compared with the direct offloading scheme,this scheme consumes 82.13% less energy and supports 52.9% more maximum task lengths.Subsequently,a multi-relay cooperative offloading scheme based on deep reinforcement learning is investigated,and an edge computing scenario consisting of a local MD,multiple RNs and multiple APs is established.The local MD can select one RN to assist in offloading part of the task segments of a computational task respectively,and the RNs can continue to offload the task segments to the MEC servers of the APs for computation.Based on this scenario,the energy consumption minimization problem under the constraint of computational latency is proposed and solved using a deep reinforcement learning approach.Simulation experiments verify the convergence of the algorithm.The multi-RN cooperative offloading scheme reduces the energy consumption by 94.15%,89.18% and 39.72% compared with the proposed benchmark schemes I,II and III,respectively.