AI-Powered Software Testing for Heterogeneous Fleet Vehicle Routing Optimization with Time Constraints

The Vehicle Routing Problem (VRP) is crucial in logistics, transportation, and distribution. Traditional VRP focuses on optimizing vehicle routes between a fixed starting point and multiple locations to minimize travel distance or time. However, these models perform inadequately in dynamic environments such as campus student path planning, which involve diverse movement patterns and time window constraints. This paper addresses campus student path planning as a Heterogeneous Fleet Vehicle Routing Problem with Time Windows (HFVRPTW) and introduces the Scooter-Aware Pathfinding with Time Windows (SAPTW) method. Students start from random points and navigate a grid-based campus to fixed destinations like dormitories and cafeterias, choosing either walking or using electric scooters available at specific locations. This study tackles key challenges including diverse movement modes, time windows for reaching destinations, automatic generation of campus maps, and random generation of student starting points and destinations. Additionally, ensuring AI-powered software testing, we developed the Grid-based Campus Map Randomized Generation (GMRG) method, a rule-based approach for creating grid maps with roads, obstacles, and specific buildings. This method provides a realistic and controlled environment for route planning tests and simulations, ensuring the robustness and reliability of the proposed solution in real-world applications. Our approach highlights the potential of integrating artificial intelligence with software testing to optimize complex routing problems with time constraints. Simulation results demonstrate that SAPTW significantly enhances student arrival efficiency, reducing average arrival time by approximately 3% to 44% compared to traditional methods.