Abstract
Research and/or Engineering Questions/Objective:
Systems have been developed which can detect a decrease in driver lane-keeping ability and notify the driver of the reduction. These systems are more effective when the time required from the decrease in lane-keeping ability until driver notification is shorter. In order to notify drivers of such a decrease more effectively, this paper studies a system which can detect a decrease of driver’s lane-keeping ability within several seconds.
Methodology:
In previous studies, because a decrease in driver lane-keeping ability was detected by statistical means based on changes in driver and vehicle behaviour, approximately 1 minute was required before the decrease in ability could be detected. To shorten this detection time, this study uses a deterministic approach for detection. The deterministic approach is based on a causal relationship, and this study uses corrective steering by the driver as the causal relationship for detection. Specifically, this system detects an interruption of the steering to correct course deviations that occurred previously. For this detection, predicted behaviour based on a driver model that was identified during the past 15 seconds is compared with the actual vehicle behaviour in real time. In particular, detection utilizes deviation from the road centre line and the target position that was predicted by an driver inverse model.
Results:
For validation tests, this system was installed in a driving simulator. The system sounds a chime when it detects a decrease in driver lane-keeping ability. To produce drowsiness in the drivers, a straight road test lasting one hour was conducted with the driving simulator. Cameras were used so that engineers who were outside the driving simulator could observe the drivers’ faces. Expression of one of the drivers indicated drowsiness 10 minutes after the test start. At that instant, the system detected a decrease in the driver lane-keeping ability and sounded the chime. The driver later reported that he had fallen asleep momentarily immediately before the chime sounded. Because the system detection results are consistent with the states reported by the drivers, we conclude that this system can detect a momentary decrease in driver lane-keeping ability within several seconds.
Limitations of this study:
The system cannot detect a decrease in lane-keeping ability even when a driver is asleep as long as the vehicle continues to follow the road. This is because the steering angle remains equal to the steering angle predicted by the driver model for the purpose of following the road.
What does the paper offer that is new in the field including in comparison to other work by the authors? :
There are two new things. One is that this system can detect a momentary decrease in driver lane-keeping ability within several seconds. The other is the method of comparing the current actual driver behaviour with the behaviour of a driver model that was identified in the past.
Conclusions:
This study examined a system that uses a deterministic approach to detect a momentary decrease in driver lane-keeping ability within several seconds. There are plans to adopt this system in new vehicles.
KEYWORDS –driver state, driver model, determinism approach, man-machine system, lane keeping