User Satisfaction-Aware Edge Computation Offloading in 5G Multi-Scenario

Abstract

Edge computation offloading has made some progress in the fifth generation mobile network (5G). However, load balancing in edge computation offloading is still a challenging problem. Meanwhile, with the continuous pursuit of low execution latency in 5G multi-scenario, the functional requirements of edge computation offloading are further exacerbated. Given the above challenges, we raise a unique edge computation offloading method in 5G multi-scenario, and consider user satisfaction. The method consists of three functional parts: offloading strategy generation, offloading strategy update, and offloading strategy optimization. First, the offloading strategy is generated by means of a deep neural network (DNN), then update the offloading strategy by updating the DNN parameters. Finally, we optimize the offloading strategy based on changes in user satisfaction. In summary, compared to existing optimization methods, our proposal can achieve performance close to the optimum. Massive simulation results indicate the latency of the execution of our method on the CPU is under 0.1 seconds while improving the average computation rate by about 10%.