top of page

Matching Rider Demand and Sharing Service in Transportation Infrastructure Networks for the Pittsburgh Metropolitan Area

Team: Sean Qian (PI, CMU),  Alex Jacquillat (Co-PI, CMU)
Funding source: Pennsylvania Infrastructure Technology Alliance
Start/End Time:  2017-2018

As of 2013, there were over 256 million private vehicles owned and operated in the United States. Estimates of the average cost to own, maintain, insure, and park a private vehicle ranges from $460 to $913 per month, an enormous economic and environmental burden. Those vehicles altogether generate on average 37 hours’ congestion per vehicle in the year of 2013. Sharing services, such as ridesharing and on­-demand taxi systems (e.g., Uber and Lyft) offer the potential to meet travel needs that are substitutional to self-­driving, which leads to significant savings of energy use and flow reduction. In turn, ridesharing can mitigate the costs, congestion and environmental impact of automobile transportation. More importantly, as they grow to represent a significant fraction of network flows, ridesharing systems can influence traffic flows in urban areas in order to improve infrastructure performance. In spite of great potential, sharing services are not provided optimally. Travelers (i.e., service demand) oftentimes have difficulty finding vehicles, or do so at high costs (e.g., Uber’s surge price). On the other hand, taxi/uber/lyft drivers (i.e., service supply) also face challenges to remain profitable, and can lose significant revenue opportunities through incomplete knowledge of traveler demand. This PITA research will address the mismatching among sharing service providers and consumers by proactively predicting traveler demand and sharing demand information to service providers, resulting in better service for travelers, better revenue opportunities for drivers, and better transportation infrastructure performance. In particular, partnering with Gridwise, a Pittsburgh-­based start­up company, the research team aim to develop predictive models from numerous input datasets that are likely to correlate with rideshare demand. The models will be trained and validated using data obtained from CMU Mobility Data Analytics Center (MAC) and the Gridwise platform, as well as other simulated “proxies” of this demand such as rideshare surge prices.

bottom of page