Real-Time Large-Scale Ridesharing with Flexible Meeting Points

In this report, the authors propose an online and large-scale rideshare system that can dynamically match passenger requests with drivers and provide efficient routes to the drivers. The authors developed a greedy insertion-based routing procedure to route thousands of requests in an hour. They incorporated flexible meeting point selection into the framework, which can reduce travel distances for both drivers and passengers. The authors implemented an online incentive and cost-sharing system that can incentivize drivers and passengers for their ride time limit violations and share the cost of a rideshare trip among the passengers fairly. The authors incorporated a request prediction and detour mechanism into the ridesharing framework. To get the most updated travel time and study the effects of ridesharing in a road network, theauthors also incorporate a simulation approach into the framework. Numerical experiments performed on the New York Taxicab dataset and a rural dataset based on Kern and Tulare Counties, California, show that the proposed framework is effective, matching thousands of requests per hour. Results also show that ridesharing can cost significantly less compared to ride-hailing services such as Uber or Lyft, and incorporating flexible meeting points can reduce travel distance by 4% on average. Simulation studies show that ridesharing can reduce total vehicle miles traveled by 13% in Manhattan on average. The proposed framework can help transportation officials design real-time and city-scale rideshare systems to alleviate traffic congestion problems in California. View the NCST Project Webpage