Genetic Algorithm Improvement: A Case Study of Capacitated Vehicle Routing Problem

Smart logistics is a crucial aspect of constructing smart cities, which entails efficiently finding a solution to a problem using a fleet of vehicles to serve geographically dispersed clients. It comprises the capacitated vehicle routing problem (CVRP), a well-known NP-hard complex optimization problem that a genetic algorithm (GA) can solve even with some weaknesses. The weaknesses include time-consuming, difficult-to-achieve convergence, and easy-to-get premature convergence, resulting in infeasible and low-quality solutions in a limited population. Based on these weaknesses, the authors propose three improvements to GA to optimize the solution. The improvement strategies are enhancing the initial population with the nearest neighbor algorithm, improving the new mutated offspring with a 2-opt heuristic, and optimizing the route with a give-and-exchange operator. The test is undergone on 53 CVRP problem sets to evaluate the performance of our proposed algorithm. The result shows that the proposed algorithm successfully improves GA performance quality, reduces the execution time, reaches some optimum values, and obtains a better solution than the best-known value.