Rapid advancements in mobile and wireless communication, cloud computing, and sophisticated artificial intelligence have precipitated technological paradigm shifts in the transportation sector. As a typical example, globally expanded app-based on-demand ride-sourcing services are revolutionizing the way of our daily mobility. In particular, ride-pooling (a common ride-sourcing service) is generally deemed to be an efficient and cost-effective way to ameliorate traffic congestion by improving vehicle utilization. Given the highly stochastic nature of traffic conditions in both space and time, we confront the challenges of precisely understanding relationships between average detour distance, average driver’s routing distance, peer-to-peer matching rate, and the different ride-pooling dem...[ Read more ]

Rapid advancements in mobile and wireless communication, cloud computing, and sophisticated artificial intelligence have precipitated technological paradigm shifts in the transportation sector. As a typical example, globally expanded app-based on-demand ride-sourcing services are revolutionizing the way of our daily mobility. In particular, ride-pooling (a common ride-sourcing service) is generally deemed to be an efficient and cost-effective way to ameliorate traffic congestion by improving vehicle utilization. Given the highly stochastic nature of traffic conditions in both space and time, we confront the challenges of precisely understanding relationships between average detour distance, average driver’s routing distance, peer-to-peer matching rate, and the different ride-pooling demand levels. To tackle these challenges, this dissertation is primarily dedicated to deepening our understanding of the intertwining relations how the aforementioned three key measurements vary with demand for shared rides in ride-sourcing markets by leveraging data-driven analytics.

We first obtain specific quantitative relations (empirical laws) calibrated from extensive numerical experiments on ride-pooling services. These experiments are underpinned by publicly available real mobility data from DiDi and New York Taxi and Limousine Commission. In addition to the above data-driven study, we then develop aggregate theoretical models by incorporating the obtained empirical laws and traffic congestion externalities to analyze the impact of ride-sourcing markets with or without ride-pooling services on average network traffic speed. The model-driven analysis allows us to figure out the optimal operating strategies of ride-sourcing platforms reacting to different congestion levels, matching time windows, and maximum matching distances for pooled rides. Our methodology and conclusions provide invaluable operational acumen for the ride-sourcing service operator and policy insights for the government/regulators. Another large chunk of this dissertation mainly investigates the impact of a newly introduced ride-sourcing service option, i.e., bundled option, which combines ride-pooling and solo ride service. Specifically, passengers who choose the bundled service option wait in the queue for both non-pooling and ride-pooling service simultaneously, and the platform assigns them either service depending on the demand, supply level, successful matching rate, inconvenience cost due to pooling, etc. Based on our models, we seek out the monopoly optimum strategy that maximizes the platform profit for heterogeneous passengers in terms of different values of time and then propose countermeasures to boost system performance.

Although this dissertation scopes out the ride-sourcing market with an emphasis on ride-pooling services, the methods, analysis, and solutions can be readily extended to address relevant research topics arising from other types of shared mobility services, e.g., multi-modal transportation or similar two-sided markets, such as bike-sharing, taxi market.