An optimization-based speed-control method for high frequency buses serving curbside stops

In this paper, an optimization-based speed-control method is proposed to alleviate the adverse effects of bus bunching. This method specifically focuses on one-way line transit corridors with curbside stops and high frequency buses. In such public transit corridors, holding buses at stops may not be suitable for implementation. Instead of holding, we choose to control bus cruising speeds between stops. The relationship between holding and speed-control is discussed. A conventional holding model M0 style= border: 0px none currentcolor; box-sizing: border-box; direction: ltr; display: inline-block; float: none; font-size: 16.2px; line-height: normal; margin: 0px; max-height: none; max-width: none; min-height: 0px; min-width: 0px; overflow-wrap: normal; padding: 0px; position: relative; white-space: nowrap; word-spacing: normal; \u3eM0 as well as a model M1 style= border: 0px none currentcolor; box-sizing: border-box; direction: ltr; display: inline-block; float: none; font-size: 16.2px; line-height: normal; margin: 0px; max-height: none; max-width: none; min-height: 0px; min-width: 0px; overflow-wrap: normal; padding: 0px; position: relative; white-space: nowrap; word-spacing: normal; \u3eM1, that determines travel times between stops, are considered. In order to make control decisions easy to understand as well as to follow, an improved speed-control model M2 style= border: 0px none currentcolor; box-sizing: border-box; direction: ltr; display: inline-block; float: none; font-size: 16.2px; line-height: normal; margin: 0px; max-height: none; max-width: none; min-height: 0px; min-width: 0px; overflow-wrap: normal; padding: 0px; position: relative; white-space: nowrap; word-spacing: normal; \u3eM2 is also proposed. All the models allow for control decisions based on the real-time state of the entire system. The three models are programmed in a rolling horizon scheme within a simulation system. Simulation experiments are conducted to examine these models in different scenarios associated with different levels of traffic randomness. The corresponding results show that, when the traffic is highly variable, the speed-control method can reduce average passenger waiting time by more than 23% style= border: 0px none currentcolor; box-sizing: border-box; direction: ltr; display: inline-block; float: none; font-size: 16.2px; line-height: normal; margin: 0px; max-height: none; max-width: none; min-height: 0px; min-width: 0px; overflow-wrap: normal; padding: 0px; position: relative; white-space: nowrap; word-spacing: normal; \u3e23% while the holding method results in an average passenger waiting time that is almost 2 min longer than when using the speed-control method. So the speed-control model is shown to be more suitable for implementation in practice. Four strategies that can generate more driver-friendly control decisions are further proposed and tested. The practical applications of the proposed speed-control models are discussed