Intelligent Buses in a Loop Service: Emergence of No-Boarding and Holding Strategies

Figure 3

Two buses with identical natural frequency serving a loop of bus stops learn the no-boarding strategy by reinforcement learning. Notably, they learn about the upper bound at some angle strictly less than to implement no-boarding, where if they exceed, then they would inadvertently implement no-boarding too frequently such that they are not meeting the level of demand for service. Also, they learn to unbunch, where one learns to stay whilst the other learns to leave, if their phase difference is .