Spatio-Temporal Capsule-based Reinforcement Learning for Mobility-on-Demand Network Coordination

As an alternative means of convenient and smart transportation, mobility-on-demand (MOD), typified by online ride-sharing and connected taxicabs, has been rapidly growing and spreading worldwide. The large volume of complex traffic and the uncertainty of market supplies/demands have made it essential for many MOD service providers to proactively dispatch vehicles towards ride-seekers.

To meet this need effectively, we propose STRide, an MOD coordination-learning mechanism reinforced spatio-temporally with capsules. We formalize the adaptive coordination of vehicles into a reinforcement learning framework. STRide incorporates spatial and temporal distributions of supplies (vehicles) and demands (ride requests), customers' preferences and other external factors. A novel spatio-temporal capsule neural network is designed to predict the provider's rewards based on MOD network states, vehicles and their dispatch actions. This way, the MOD platform adapts itself to the supply-demand dynamics with the best potential rewards. We have conducted extensive data analytics and experimental evaluation with three large-scale datasets (~ 21 million rides from Uber, Yellow Taxis and Didi). STRide is shown to outperform state-of-the-arts, substantially reducing request-rejection rate and passenger waiting time, and also increasing the service provider's profits, often making 30% improvement over state-of-the-arts.