Modeling Driver Adaptation Capabilities in Critical Driving Situations 2012-01-1179
In order to improve driver safety, car manufacturers propose a wide range of Advanced Driving Assistance Systems (ADAS). Their behavior can drastically modify the vehicle dynamics and generate for the driver a discrepancy between the expected and observed behavior. The modification of car response can also be due to the occurrence of an ADAS component failure. It appears the assessment of the impact of such systems and such events on the driver is a major challenge, which designers have to cope with. However, it appears, the monitoring of driver's response may become complex to implement on real car or even hazardous. The introduction of driving simulator allows overcoming many of these limitations but the conduct of exhaustive experiments is also difficult considering the number of possible events, driving context and involving a wide range of drivers' profiles.
The use of driver models which is able to reflect driver control behavior in such conditions is necessary. The driver adaptation to inexperienced and sudden environment changes can involve two different strategies. The first one, the impedance modulation, is due to neuromuscular properties: the increase of muscles stiffness allows rejecting external disturbance on manipulated objects such as the steering wheel. The second one is the learning and the update of the driver internal representation of the car dynamics. In this study, we propose a driver model integrating both these strategies. The implementation is inspired of classical adaptive control theories: high gain adaptive control for impedance modulation and model reference adaptive control for internal representation update. As an illustration, we consider the case of a sudden and permanent increase of torque level on the steering wheel during a curve driving. At last, we propose how to use this model for safety studies and Automotive Safety Integrity Level (ASIL) rating.