Dynamic Taxi Ride Sharing using Localized Communication

With the rise of on-demand taxi services, like Uber, Lyft, etc., urban public transportation started to heavily depend on taxicabs. While increasing demand leads to passenger stranding, higher supply may result in traffic congestion. In order to strike a balance ride sharing plays an important role. However, scheduling a ride on-the-fly is extremely challenging and more so with increasing number of passengers. It is non-trivial to find a driving route for the taxi accommodating multiple passengers without extending their journey time beyond a pre-specified tolerance value. The spatiotemporal separation of passengers and high mobility of taxis even more complicates shared ride scheduling. Existing distributed ride sharing solutions failed to address the extremely dynamic nature of the underlying topology where taxis are continuously moving and hence, results in message loss. In this paper, we have proposed a purely distributed ride sharing algorithm aimed at addressing the dynamics of taxi topology using asynchronous localized communication between passengers and taxis. Empirical analysis using large scale single-user taxi ride records from Chicago, show that, our proposed algorithm, ensures a maximum of 76% success in ride sharing and a 97.5% taxi occupancy rate during peak operating hours.