A Reinforced Neighborhood Search Method Combined With Genetic Algorithm for Multi-Objective Multi-Robot Transportation System

With the rapid advancement of artificial intelligence, autonomous multi-robot systems have been successfully applied to various domains. Therefore, developing intelligent routing and scheduling systems to efficiently coordinate multi-robot movements in transportation networks emerges as a critical challenge. To address this issue, this study constructs an optimization model for cooperative robot operations, aiming to minimize total energy consumption and the completion time of most time-consuming robot. These objectives contain conflicts, thus requiring a multi-objective optimization approach to resolve them. We propose a reinforced neighborhood search method combined with genetic algorithm (RNSGA), which combines single solution search ideas and population-based techniques. RNSGA consists of two crucial steps: route construction to determine the composition and visiting sequence of task points within each route, as well as route allocation to assign routes to individual robots. The route construction phase incorporates several key components, including solution initialization, route balance mechanism, proximity-based optimization mechanism, and intro-route sequence adjustment method. For the route allocation phase, a population-based allocation mechanism is employed to determine the optimal assignment of routes. Comprehensive experiments on 24 classic transportation test instances demonstrate that RNSGA significantly outperforms six state-of-the-art algorithms.