A human driver is expected to make multiple decisions and / or to perform multiple operations in parallel when he/she is driving a car. The driver has to track the desired road path while following the speed limit, and put appropriate distance with other cars and pedestrians to avoid collisions. The driver has also to cope with the changes of car response characteristics which might be caused by variations of weather, number of passengers and road conditions. A number of researches have been conducted to establish a mathematical model which describes the control action of a human driver under the context of model predictive control theory. This paper handles the same problem of modeling control action of a human subject engaged with a multiobjective control task. A two dimensional tiny driving simulator has been developed to analyze control action of the human subject trying to accomplish designated task under several different scenarios. Proposed model is a linear regression model where extrapolated physical variables expressing relations with other cars are incorporated as quantities related to control action of a human subject. Analysis of manual control action while he/she was operating a 2D driving simulator to complete the assigned task indicates that the proposed model not only provides an inference on how human control action is synthesized but also claries how much manual control action would rely on prediction might vary depending on the operator's experience and control scenarios.
Pilot model
Manual predictive control
Regression analysis
Principal component analysis