Multi-Objective Resource Allocation in Edge Computing Using Improved Genetic Algorithm with Knowledge-Based Crossover and Segmentation Mutation

Abstract

With the widespread adoption of edge computing technology, efficient resource allocation becomes an urgent problem to be solved. Aiming at the optimization of resource allocation in the edge computing environment, a multi-objective optimization function with three dimensions of time, energy consumption, and computing resource occupation is constructed. An improved genetic algorithm resource allocation strategy model based on a knowledge-guided cross-segmentation mutation mechanism is proposed. Knowledge crossover enhances the global search capability by prioritizing the retention of gene segments with higher historical fitness contributions. The experimental results showed that the loss function value of the new model could be as low as 0.012 during the training process, indicating that the model convergence on multi-objective optimization was effectively improved. Meanwhile, the improved genetic algorithm model generated an average of 4.8 optimal solutions, which was 1.8 more than the traditional genetic algorithm which generated an average of 3.0 optimal solutions. Compared with multi-objective optimization algorithms such as NSGA-II, SPEA, and MOPSO, the research model reduced the average energy consumption of the device to 112.68 Joules and the average energy consumption of the system to 208.12 Joules in the EdgeDroid dataset. Its utilization of computational resources reached 77.45%, with the processing time of the task shortened to 5.44 seconds. In real application scenarios, the model achieved a 92.1% and 89.2% task completion rate in shopping malls and hospital environments, and the resource utilization rate was improved to 83.49% and 81.67%. In summary, the proposed method effectively improves resource allocation efficiency, reduces system energy consumption, and optimizes task completion performance, which provides strong support for dynamic resource management in edge computing environments.