A Graph Matching Based Method for Dynamic Passenger-Centered Ridesharing

摘要

Ridesharing is one transportation service deeply influenced by the prosperity of Mobile Internet. Existing work focuses on passenger-vehicle matching, which considers how to optimally dispatch passengers to appropriate vehicles. While dynamic passenger-passenger matching addresses how to optimally handle continually-arriving requests for ridesharing from passengers, without considering vehicles. It is a kind of dynamic passenger-centered ridesharing that has not been studied enough. This paper studies dynamic passenger-centered ridesharing with both temporal and cost constraints. We first propose a ridesharing request matching method based on maximum weighted matching on undirected weighted graphs, aiming to minimize the overall travel distance of targeting passengers. We then devise a distance indexing strategy to prune unnecessary calculations to accelerate ridesharing request matching and reduce request response time. Experiments on real-life road networks indicate that our method can successfully match 90% of ridesharing requests while saving 23% to 35% of travel distance.