Stochastic model predictive control with driver behavior learning for improved powertrain control

• Published in: 49th IEEE Conference on Decision and Control (CDC) pp. 6077 - 6082
• Main Authors: Bichi, M, Ripaccioli, G, Di Cairano, S, Bernardini, D, Bemporad, A, Kolmanovsky, I V
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2010
• Subjects: Adaptation model; Batteries; Driver circuits; Markov processes; Predictive models; Vehicles
• ISBN: 142447745X; 9781424477456
• ISSN: 0191-2216
• DOI: 10.1109/CDC.2010.5717791

In this paper we advocate the use of stochastic model predictive control (SMPC) for improving the performance of powertrain control algorithms, by optimally controlling the complex system composed of driver and vehicle. While the powertrain is modeled as the deterministic component of the dynamics, the driver behavior is represented as a stochastic system which affects the vehicle dynamics. Since stochastic MPC is based on online numerical optimization, the driver model can be learned online, hence allowing the control algorithm to adapt to different drivers and drivers' behaviors. The proposed technique is evaluated in two applications: adaptive cruise control, where the driver behavioral model is used to predict the leading vehicle dynamics, and series hybrid electric vehicle (SHEV) energy management, where the driver model is used to predict the future power requests.