Experimental Evaluation of a Game-Theoretic Human Driver Steering Control Model

Area of Science:

• Automotive Engineering
• Control Systems Theory
• Human-Machine Interaction

Background:

• Automated vehicle steering control systems offer significant road safety benefits.
• Accurate mathematical driver models are crucial for developing effective automated steering interventions.
• Existing models may not fully capture human steering responses to automation.

Purpose of the Study:

• To experimentally evaluate a novel game-theoretic driver steering control model.
• To compare the performance of the game-theoretic model against a conventional optimal-control-theoretic model.
• To assess the models' ability to represent real human steering behavior during automated steering intervention.

Main Methods:

• Developed a game-theoretic driver model based on Nash equilibrium for driver-automated controller interaction.
• Identified model parameters by fitting to steering behavior data from six human subjects in a driving simulator experiment.
• Evaluated model performance by comparing fitting errors against a conventional optimal-control-theoretic driver model.

Main Results:

• The game-theoretic driver model demonstrated statistically significant superiority in representing the steering behavior of three out of six subjects.
• For the remaining three subjects, both the game-theoretic and conventional models performed with statistically equivalent accuracy.
• The study provides empirical evidence for the effectiveness of game-theoretic approaches in driver modeling.

Conclusions:

• The game-theoretic driver steering model offers a more accurate representation of human steering behavior compared to conventional models for a subset of drivers.
• This improved accuracy has implications for the design and implementation of safer and more intuitive automated vehicle systems.
• Further research can explore the factors influencing individual differences in model performance and refine game-theoretic strategies for human-automation interaction.