UC Irvine

The possibilities of enabling travelers to walk a short distance to reach their prescribed pickup locations or final destinations in ridesharing services have received a lot of attention in recent years. This thesis classifies the existing literature on ridesharing with walking trip legs into three different classes. Additionally, this thesis presents a variant of the single-vehicle pickup and delivery problem with time windows (PDPTW), called the pickup and delivery problem with time windows and walking legs (PDPTWWL), that explicitly incorporates the access and egress walking trips of travelers into the decision space. The proposed model targets a ridesharing service that can pool together two to three requests, and it is solved to optimality using a commercial mixed integer programming (MIP) solver. The candidate pickup/drop-off (PUDO) locations for each request are either the road intersections or street centerline midpoints within a pre-defined maximum allowable walking time. The study applies the model to two road network datasets—Isla Vista and Chicago downtown—reflecting different real-world application scenarios, and it presents various performance metrics based on a large number of randomly generated 2-request, 3-request, and 4-requesst PDPTWWL instances. The results indicate that a relatively short allowable walking time has a comparatively high vehicle driving time reduction per second of walking (DTRPSW) relative to a long allowable walking time. Moreover, pooling a larger number of requests (e.g., four requests) in one vehicle can lead to a much larger reduction in vehicle driving time and a higher DTRPSW, compared to pooling only two requests in one vehicle.