Multi-target Longitudinal Control Based on Model Predictive Control for Autonomous Bus

In this paper, a framework for the optimal longitudinal control of autonomous buses is proposed for multi-target scenarios. Autonomous buses operate on roads that present various events, such as vehicular interactions, traffic signals, and bus stop pauses. Consequently, the development of suitable acceleration/deceleration strategies and control systems that effectively respond to each situation is essential. For each event, a reference acceleration model is established. Priorities amongst numerous events are ascertained by comparing the reference accelerations designed for each situation. As for the longitudinal controller, a novel bus model is developed to embody the unique actuator characteristics inherent to large buses. Moreover, Model Predictive Control (MPC) is utilized to determine the optimal longitudinal acceleration for the selected target, enhancing passenger comfort by limiting acceleration and minimizing actuator transitions. The proposed longitudinal control system is validated through field tests in a complex road environment.