Pricing and Allocation Algorithm Designs in Dynamic Ridesharing System

Chaoli Zhang,Jiapeng Xie,Fan Wu,Xiaofeng Gao,Guihai Chen

DOI: https://doi.org/10.1016/j.tcs.2019.05.045

IF: 1.002

2019-01-01

Theoretical Computer Science

Abstract:Traffic congestion and car pollution are becoming serious plagues nowadays. High travel cost brings a great burden to people and society. A ridesharing system mitigates traffic congestion and car pollution by allowing passengers to share their travel costs with others. Traditional ridesharing platforms usually require passengers to submit their plans in advance and then design schedules for the drivers who would like to offer a ride. Nowadays, with the development of the smartphone technology, dynamic ridesharing systems enable passengers request a car anytime and anywhere. This paper mainly considers the problems of how to allocate passengers to drivers, how to charge the passengers and how to design feasible schedules for the driver in such online environment. The allocation problem is modeled as an online weighted matching problem with the graph changing over time. Firstly, we give a fair pricing method which is easy to be understood and accepted by the passengers. We develop a greedy algorithm called LiqMax_Gre for the purpose of maximizing liquidity and an algorithm called UtiMax for the purpose of maximizing utility. LiqMax_Gre achieves a competitive ratio of 1 λ + 1, where λ is the maximal number of passengers a car can take. In general, we prove that no online algorithm can have a good guarantee for the design goal of maximizing utility. Innovatively, UtiMax considers not only the current maximal utility, but also the opportunity cost, which is the utility contributed by occupying a seat. It reflects the utility in the future and thus can be used to handle the difficulty in online design. We prove that, our algorithm has a competitive ratio of 1/3 in a special case. The schedule problem is NP-hard and we design a heuristic nearest neighbor algorithm to solve it.