Ride-sharing is About Agreeing on a Destination

Ride-sharing is rapidly becoming an alternative form of transportation mainly due to its economic benefits. Existing research on ridesharing aims to optimally match trajectories between people with pre-selected destinations. In this paper, we show better ride-sharing arrangements are possible when users are presented with more destinations and agree on a common destination. Given a set of points of interest (POIs) and a set of users, our approach presents destination POIs and computes ride-sharing plans. Each arrangement for a subset of users that fit in a car can be presented as a minimum Steiner tree (MST) problem. An optimal solution of the overall problem minimizes the total length of all the MSTs. The problem is a version of the set cover problem and is NP-hard. We first develop a series of baseline methods which use a popular MST algorithm. Then, we propose our method which uses constraints on intermediary points where users can meet to share rides. These constraints reduce the time complexity significantly and our method is up to two orders of magnitude faster than the best baseline method. Since our algorithm finds the subsets of users and POIs for each arrangement, we define and solve a new type of MST problem as a first step. Our experiments show that our method can provide a fast and readily deployable solution for real world large city scenarios.