Sign up for our newsletter
Technology / AI and automation

PREDICTING DRIVER ALERTNESS FROM STEERING BEHAVIOUR

Research into predicting driver alertnesss from steering behaviour is being carried out by the University of Stirling in conjunction with Ford Motor UK Ltd. The project started about 18 months ago and the idea was to develop some kind of neural network-based system that could be fitted into a car to determine when the driver is getting drowsy, but does not allow the situation to go as far as waking them up if they have dropped off. A test car was fitted out to give a whole series of medical measures on the driver’s alertness, and pressure on the steering wheel proved to be the most practical in terms of training a neural network, and was found to be closely linked with the alertness of the driver. Runs of between two and four hours down a motorway were carried out at various times of day to collect the data, with some drivers being sleep-deprived the night before. For safety there was also a co-driver. Forty different drivers did motorway runs and the neural network was trained from 15 of those. A feed forward backpropagation multilayer perceptron network was used. Results showed that the neural network made errors in predicting that the driver was falling asleep where there was original doubt about the data, which pointed to a problem with the training set used on the network rather than with the system itself.

White papers from our partners


This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.