Intelligent Carpool Routing for Urban Ridesharing by Mining GPS Trajectories

Abstract

To support an efficient carpooling service in heavy urban traffic, we propose an intelligent routing scheme based on mining Global Position System trajectories from shared riders. The carpooling system provides many-to-many services with multiple pickup and dropping points. To join a daily carpooling group, the riders must accept a compromised route that is efficient after merging the routes that are preferred by all qualified riders. We developed three frequency-correlated algorithms for route mining, rider selection, and route merging in an urban carpool service. Our approach can cope with the traffic dynamics to yield a suboptimal shared route. Our scheme was successfully tested under heavy Beijing traffic over hundreds of riders. We developed performance metrics to measure the service cost and mileage saved. The ultimate goal is to minimize the riding distances and the transportation costs, and thus alleviate urban traffic jams.