Deep Reinforcement Learning based dynamic optimization of bus timetable

Bus timetable optimization is a key issue to reduce operational cost of bus company and improve the transit service quality. Existing methods optimize offline. However, in practice, short-term passenger flow may change dramatically from time time. Timetables generated offline cannot be adjusted real handle changed flow. In this paper, we propose Deep Reinforcement Learning based Timetable dynamic Optimization method (DRL-TO). DRL-TO, problem formulated as Markov Decision Process (MDP). A Q-Network (DQN) applied agent decide whether departs at each minute during period. Therefore, departure intervals services are determined accordance with demand. We identify several new useful state features for DQN agent, including load factor, carrying capacity utilization rate, passengers’ waiting number stranded passengers. Considering interests both passengers, reward function designed, which includes metrics full empty time, Experiments demonstrate that, comparison by approaches manual method, DRL-TO can dynamically determine on real-time flow, generate less points (i.e., cost) shorter higher quality service).