Optimal Vehicle Dispatching Schemes via Dynamic Pricing
share
Recently, shared mobility has been proven to be an effective way to relieve urban traffic congestion and reduce energy consumption. Despite the emergence of several nationwide platforms, the pricing schemes and the vehicle dispatching problem of such platforms are optimized in an ad hoc manner. In this paper, we introduce a general framework that incorporates geographic information and time-sensitive dynamic environment parameters (such as the dynamically changing demand) and models the pricing and dispatching problem as a Markov Decision Process with continuous state and action spaces. Despite of the PSPACE-hardness of general MDPs, we provide efficient algorithms finding the exact revenue (or welfare) optimal (potentially randomized) pricing schemes. We also characterize the optimal solution via primal-dual analysis of a convex program. Finally, we also discuss generalizing our model by showing how to reduce a wide range of general settings in practice to our model.
READ FULL TEXT