Because existing public transportation infrastructure cannot be adapted in a timely manner to address the daunting traffic and parking congestion in urban environments, researchers are investigating social solutions, such as carpooling, where a driver and one or more passengers having semi-common routes share a private vehicle. Although many carpooling systems have been proposed, most of them lack various levels of automation, functionality, practicality, and solution quality. While Genetic Algorithms (GAs) have been successfully adopted for solving combinatorial optimization problems, their use is highly uncommon in carpooling problems. Motivated to propose a solution for the many to many carpooling scenario, we present in this paper a GA with a customized fitness function that searches for the solution with minimal travel distance, efficient ride matching, timely arrival, and maximum fairness while taking into account the riding preferences of the carpoolers. The computational results and simulations based on real user data show the merits of the proposed method and motivate follow up research.

No abstract