A Hyper-Heuristic Approach to Solving Vehicle Routing Problem in Military Logistics Distribution

Today’s war tactics and strategies are planned for the use of smart guns (artificial intelligence supported) rather than human power. With the “Information Age” or the “Digital Age”, military software has developed in the following areas; developed new guns, land warfare systems, communication systems, advanced optics and lasers, image processing systems, artificial intelligence supported autonomous or non-autonomous robot technologies, etc. Developments in these areas significantly improve the security systems of countries and cause different countries to develop advanced strategies instead of traditional war strategies. In this study, a vehicle routing and depot management system was developed to enable easy procurement of material for soldiers based on the material (guns, explosives, food-beverage, health, etc.) shortage they encounter during operations. Google Maps API was used for mapping and determining the soldier’s location, RoboFlow API was applied for learning and processing the physical conditions of the current location, and C# along with Microsoft SQL Server program was used for the development of the web application. The application consists of two stages. In the first stage of the application, an aerial vehicle was used to determine the geographical location of the soldier and the physical conditions of the location (mountainous, forested, road, water, etc.). The obtained images were transferred to the web application developed in the ASP.NET environment. These transferred images are processed by artificial intelligence trained by supervised learning using the YOLOv8 architecture via the RoboFlow API. In this way, the physical conditions of the soldier’s location are classified and defined. In the second stage of the application, the information on the materials needed is obtained 2 from the user and the best routes are created using genetic algorithms and tabu search algorithms (hyper-heuristic) according to the stock amounts in the depots. Thanks to this developed application, the results obtained using the hyper-heuristic algorithm were compared with the results obtained without using the hyper-heuristic algorithm, and it was observed that the running time of the program and the number of iterations decreased. It is thought to focus on machine learning and explainable artificial intelligence in future studies.