Research on transfer discount for public transportation based on deep reinforcement learning and game theory

Reasonable transit transfer discount policies incentivize people to take public transportation and alleviate the increase in vehicle usage. This research establishes and analyzes a dynamic game model with perfect information for public transportation operator and travelers as players. By treating the optimization process of discount as a learning process for operator to continuously explore and obtain the optimal price in a complex traveler environment, an adaptive discount adjustment method based on deep reinforcement learning is proposed. The result shows that transfer discount can increase the travel share of public transportation; Operator can use the first mover advantage to and achieve the most favorable Nash equilibrium for himself; The applicability of the proposed method in transfer discount scenarios is tested via simulation; According to the reinforcement learning DDPG algorithm, the Nash equilibrium is reached when the fare change interval is between (-0.68, -0.52) yuan, resulting in an increase in the usage of public transportation.