Exploring the Impact of Car-Following Models and Controller Formulations on Autonomous Vehicle Motion Planning

Abstract

Effectively planning the behavior of autonomous vehicles (AVs) while accounting for their mechanical properties and traffic flow characteristics is a challenging task. This study evaluates different car-following models in combination with various controllers to model the longitudinal behavior of AVs. Controllers regulate a vehicle's speed, ensuring smooth acceleration in accordance with its planned path. Specifically, conventional models, interaction-based models, and artificial intelligence-based models were tested alongside standard controllers. The respective transfer functions of the system were derived, and the weights were tuned accordingly. Nanoscopic simulation runs were conducted to assess performance. The results indicate that the choice of car-following model and controller significantly impacts the longitudinal planning of AVs, with certain combinations demonstrating superior performance. This study thus provides a framework for identifying the optimal pairing of a car-following model and controller to enhance longitudinal behavior planning in AVs.