A Driver-Adaptive Warning System (DAWS) applies learning algorithms to develop individualized models of a human driver's 'style' and uses these models to tailor warnings to be more appropriate and useful for the individual. Alarms will be state-based, provided when the driver's current state indicates a potentially dangerous situation, or event-based, provided when a current event poses imminent danger to the driver. Of the methods tried, the best performance was obtained using a multi-layered perceptron neural network approach with three layers of nodes operating over five inputs. Such networks were 25% better at predicting the future lane following behaviors for the individual than for other drivers in the sample. As there are significant individual differences in driving style, a DAWS holds promise for increasing driver acceptance and use, and thus driver safety.
Driver-Adaptive Warning System
1995
22 pages
Report
Keine Angabe
Englisch
Transportation Safety , Road Transportation , Transportation & Traffic Planning , Transportation , Human Factors Engineering , Traffic safety , Driver behavior , Warning systems , Adaptive control systems , Accident prevention , Models , Algorithms , Human factors engineering , Monitoring , Neural networks