Abstract
Reliable knowledge about specific characteristics of the driver, such as his current performance level, his feedforward and feedback control behaviour or his individual driving style, contains large potential for the adaptation of advanced driver assistance systems and the optimization of customer benefit (safety and comfort gain) as well as customer acceptance.
At the Institute of Automotive Engineering (IAE), a method was developed which allows the monitoring of different driver characteristics by determining defined objectification parameters which result from the systematic analysis of 3D (Driver, Driven Vehicle and Driving Environs) interaction mechanisms. The underlying database can be subdivided into test series which model real customer behaviour according to the 3D-interaction and test series for data acquisition for the identification of the driver´s condition.
In conjunction with the objectification of driving style and driver’s performance level (DPL), a control theoretical driver model is introduced. The use of the control theoretical driver model in terms of different evaluation objectives will be illustrated by means of different ways of application. The first case aims at the aspects of adaptation mechanisms to altered vehicle dynamics and driving situations and how these can be used to represent valid objective evaluation criteria for driving properties (e.g. related to age and experience) by using the driver model parameters. The second case aims at objectifying the driver in the overall system. This is based on identifiable changes in driver behaviour in relation to driving time to e.g. quantify the decreasing quality of the driver controlling the vehicle. Thus the second case aims on the identification of driver performance and alertness index through sensor information from track monitoring systems.
The individual fingerprint and the time-variant characteristics of the driver form the basis for the design of adaptive assistance systems. The first two categories (driving style and driving strategy) are used primarily for the adaptation of the basic characteristics. The continuous observation of the driver´s performance level over the driving time serves to adapt thresholds of interference for the transition from warning to intervening assistance functions.
Finally, the field applications of objectifications based on the driver’s fingerprint will be presented, supported through a number of test results. The applications cover different ways of adapting ADAS-functions as well in longitudinal as in lateral control. A practical example (cf. figure 1) shows IAE ADAS design of an adaptive lane keeping system with individual mechanisms for the adaptation of additional steering wheel torques as well as warning and interference thresholds for the transition between different intervention steps. Driver-specific acceptance and subjective safety gain can be increased significantly as compared with the reference system.
For additional tests of adaptive assistance functions and in order to ensure them for a wide range of customers, the current research steps include comprehensive system studies. In addition, the potential application of these approaches to intervening systems for longitudinal control is being studied.
Keywords: Adaptive ADAS, Driver Model, Driver’s Fingerprint, Objectification