SACA: An End-to-End Method for Dispatching, Routing, and Pricing of Online Bus-Booking

Yucen Gao, Yulong Song, Xikai Wei, Xiaofeng Gao, Guihai Chen

Abstract: Bus-booking is a novel service proposed to handle the massive traffic demand. Passengers can reserve bus seats through a bus-booking platform and it will place multiple passengers on one bus with a customized route. Order dispatching, route planning, and trajectory pricing are three key problems for this service, which interact intimately and are inappropriate to be solved separately in multi-stage. Consequently, we propose a Soft Actor-Critic (SAC) and attention based framework called SACA to solve these three problems jointly in an end-to-end manner. With the solutions of order dispatching and route planning obtained by an attention mechanism, SACA determines the order price for higher future revenue by a SAC network. SACA can deal with online scenarios and through various experiments based on real-world data, we prove its effectiveness compared with other multi-stage methods.