Predictive cruise control of connected and autonomous vehicles via reinforcement learning.

Weinan Gao; Odekunle, Adedapo; Yunfeng Chen; Zhong-Ping Jiang

Predictive cruise control concerns designing controllers for autonomous vehicles using the broadcasted information from the traffic lights such that the idle time around the intersection can be reduced. This study proposes a novel adaptive optimal control approach based on reinforcement learning to solve the predictive cruise control problem of a platoon of connected and autonomous vehicles. First, the reference velocity is determined for each autonomous vehicle in the platoon. Second, a data-driven adaptive optimal control algorithm is developed to estimate the gains of the desired distributed optimal controllers without the exact knowledge of system dynamics. The obtained controller is able to regulate the headway, velocity, and acceleration of each vehicle in a suboptimal sense. The goal of trip time reduction is achieved without compromising vehicle safety and passenger comfort. Numerical simulations are presented to validate the efficacy of the proposed methodology.

CRUISE control; AUTONOMOUS vehicles; REINFORCEMENT learning; ADAPTIVE control systems; ROAD interchanges & intersections; SYSTEM dynamics

IET Control Theory & Applications (Wiley-Blackwell), 2019, Vol 13, Issue 17, p2849

1751-8644

Academic Journal

10.1049/iet-cta.2018.6031