A scalable non-myopic dynamic dial-a-ride and pricing problem for competitive on-demand mobility systems

We propose a competitive on-demand mobility model using a multi-server queue system under infinite-horizon look-ahead. The proposed approach includes a novel dynamic optimization algorithm which employs a Markov decision process (MDP) and provides opportunities to revolutionize conventional transit services that are plagued by high cost, low ridership, and general inefficiency, particularly in disadvantaged communities and low-income areas. We use this model to study the implications it has for such services and investigate whether it has a distinct cost advantage and operational improvement. We develop a dynamic pricing scheme that utilizes a balking rule that incorporates socially efficient level and the revenue-maximizing price, and an equilibrium-joining threshold obtained by imposing a toll on the customers who join the system. Results of numerical simulations based on actual New York City taxicab data indicate that a competitive on-demand mobility system supported by the proposed model increases the social welfare by up to 37% on average compared to the single-server queuing system. The study offers a novel design scheme and supporting tools for more effective budget/resource allocation, planning, and operation management of flexible transit systems.