Solving the Vehicle Routing Problem via Distance-Aware Clustering and Simulated Annealing

The problem of routing a given number of vehicles leaving a depot to serve customers is known as the Vehicle Routing Problem (VRP). VRP is used in various fields such as logistics, supply chain and distribution. To solve VRP, in this study, we propose a solution which uses a heuristic algorithm that we developed to distribute customers to vehicles and then optimizes the route of each vehicle using Simulated Annealing technique. Our solution aims to solve VRP by generating routes of similar length for each vehicle in a short enough time to be used in real-time applications when no capacity value is given for the vehicles. To measure the performance of our solution, we compared it with OR-Tools, an open source VRP library, using problem instances that we have created by generating synthetic data. We found that in most cases it performed better and was able to create shorter routes. Thus, we consider it as an effective and performant solution for classical VRP. Since we offer a direction-oriented solution, we think that it produces useful routes in reallife problems, especially in distribution-based real-life problems.