Implementation and Analysis of Multi-Objective Task Unloading Optimization Algorithm for Vehicular Edge Computing Load Balance

In the Vehicular Edge Computing (VEC) system, the task offloading decision for delay optimization can minimize the total time consumption between end users and edge server during the offloading process. Consider the mobility of vehicular users, developing appropriate task offloading strategies poses certain challenges. Due to the characteristics of vehicle units gathering and disbanding in a short period of time in the VEC service coverage, a large number of vehicle tasks offloading causes the overload of nearest MEC servers. In this research, we implement the Multi-objective Task Unloading Optimization (MTUOA-LP) algorithm, which includes Load Prediction based on LSTM Neural Network Model using Adaptive Genetic Algorithm (LP-AGA-LS) and Multi-objective Task Unloading Optimization based on Non-dominated Sorting Genetic Algorithm III to perform load balancing of VEC servers. The LP-AGA-LS establishes a load prediction algorithm of computing resources for VEC servers in order to improve the accuracy of computing resource prediction. And the LP-AGA-LS designs a task offloading strategy with the goal of time delay with VEC server load balancing to construct a multi-objective optimization model. Based on our experimental results, we can say that the MTUOA-LP algorithm can optimize task latency, achieve significant advantages in load balancing, and solve the problem of imbalanced load on MEC servers.