Designing a Forward-looking Matching Policy for Dynamic Ridepooling Service
The fast development of information technologies in recent decade has greatly facilitated large-scale implementation of dynamic ridepooing services, e.g., Uber Pool, Didi Pinche. In dynamic
ridepooling services, service providers respond to on-demand mobility requests immediately, dispatch (vacant or partially occupied) vehicles in real-time, and keep searching for matching
orders along the trip. Most existing dispatching strategies ignore forthcoming matching opportunities, therefore having short-sighted limitations. In this paper, we argue that the system
may benefit from strategically giving up certain current matches, and propose a probabilistic matching policy under which appeared matching opportunities are accepted with varying
probabilities. Assuming that each ridepooling passenger shares vehicle space with at most one another during the entire trip, and ridepooling orders between each OD pair appear following
a Poisson process with a given rate in each study period, we propose a system of nonlinear equations to predict the system performance and the matching potential of each OD pair under
any probabilistic matching policy. Based on the model, we then propose an efficient solution algorithm to optimize the probabilistic matching policy (i.e., the acceptance probability of each
match) for minimal expected total ride distance per unit period. The optimized probabilistic matching policy allows us to make decisions encompassing a consideration of all potential
matches during each trip; therefore, it has a forward-looking feature. Through simulation experiments conducted on grid networks and the real road network of Haikou (China) utilizing a real order dataset, we demonstrate that our model yields accurate predictions of the average ride/shared distance for each origin-destination (OD) pair across various matching policies. Furthermore, the optimal matching policy generated by our method can reduce the average total ride distance per unit period by over 5% when demand is high.
Dr. Xiaolei Wang is a Professor in the School of Economics and Management, Tongji University. She got her Bachelor’s degree from the University of Science and Technology of China (USTC) in 2008, and her PhD from the Hong Kong University of Science and Technology (HKUST) in 2012. During her study at USTC and HKUST, she received the Guo Moruo Scholarship and the HKUST SENG PhD Research Excellence Award. Her major research area covers the modeling, analysis and management of the urban transportation system in the era of shared mobility and the operations research for different types of shared mobility services. She has authored or coauthored 20+ papers in peer-reviewed journals, including 8 in Transportation Research Part B and 3 in Transportation Science. The average citation of her work is 70, and her H index is 17 according to Google Scholar. She has been the PI of 4 NSFC projects. In 2020, she received the NSFC support for excellent young scholars.
3 October 2023 (Tuesday)
Professor Xiaolei Wang