Bias-free identification of a linear model-predictive steering controller from measured driver steering behavior

Area of Science:

• Automotive Engineering
• Control Systems
• Human-Machine Interaction

Background:

• Driver steering control modeling is advancing, but formal system identification methods are underutilized.
• Existing models often lack rigorous validation with experimental driver data.

Purpose of the Study:

• To develop and validate a steering controller using linear model-predictive control (MPC).
• To apply indirect system identification to minimize steering angle prediction error.
• To address identification bias in closed-loop driver-vehicle systems.

Main Methods:

• Linear model-predictive control (MPC) for steering control.
• Indirect system identification minimizing steering angle prediction error.
• Prediction error filtering to mitigate closed-loop bias.
• Application to data from 14 drivers in double lane change maneuvers.

Main Results:

• The identification procedure successfully identified model parameters yielding small prediction errors.
• The method effectively distinguished between different driver steering strategies.
• Validated the efficacy of MPC and system identification in driver steering modeling.

Conclusions:

• Formal system identification methods can be rigorously applied to driver steering control.
• The developed method provides accurate driver behavior models and differentiates driving styles.
• This approach enhances understanding and modeling of driver-vehicle interaction.