Implementation of SUMO Simulation for Comparison of CVRP

With the rapid increase in human density, development, and mobility in urban areas, the need for logistics distribution systems is increasing which is an important part of connecting industries with their consumers. Thus, route planning is an important thing for the industry. Therefore, this paper proposes a comparison of several vehicle routing problems algorithms and test the routes that have been obtained on a simulation system based on real conditions. Our proposed algorithm consists of Mixed Integer Linear Programming (MILP), Clarke-Wright and Reinforcement Learning algorithm using Markov Decision Process. Digital maps, customer data and route planning results will be converted into a SUMO simulation. We compare the performance of the algorithm with parameters consisting of the number of routes, distance traveled, computation time and simulation time. The experimental results show that the MILP algorithm has the best performance with the most optimal route results, but other algorithms have a lower computation time.