Algorithms for the Ridesharing with Profit Constraint Problem

Mobility-on-demand (MoD) ridesharing is a promising way to improve the occupancy rate of personal vehicles and reduce traffic congestion and emissions. Maximizing the number of passengers served and maximizing a profit target are major optimization goals in MoD ridesharing. We study the ridesharing with profit constraint problem (labeled as RPC) which considers both optimization goals altogether: maximize the total number of passengers subject to an overall drivers' profit target. We give a mathematical formulation for the RPC problem. We present a polynomial-time exact algorithm framework (including two practical implementations of the algorithm) and a (1/2)-approximation algorithm for the case that each vehicle serves at most one passenger. We propose a (2/3*lambda)-approximation algorithm for the case that each vehicle serves at most lambda >= 2 passengers. Our algorithms revolve around the idea of maximum cardinality matching in bipartite graphs and hypergraphs (set packing) with general edge weight. Based on a real-world ridesharing dataset in Chicago City and price schemes of Uber, we conduct an extensive empirical study on our model and algorithms. Experimental results show that practical price schemes can be incorporated into our model, our exact algorithms are efficient, and our approximation algorithms achieve about 90% of optimal solutions, in the number of passengers served.