The On–Line Use of Performance Measures to Predict Driving Fatigue

Abstract

The experiment reported herein examined the use of driving performance measures to predict whether a driver was suffering from sleep-loss and/or excessive driving time. The experiment was conducted on a closed 7.2 km oval track. Each subject driver drove around the track at about 80 kph for three hours on the first test day. For the next 21 hours, each driver remained awake and, at regular selected intervals, performed a battery of behavioural tests. On day two, 24 hours after beginning the first driving period, each subject began to drive for a second 3-hour period. The data from each driving period in the first 27-hour session were analyzed to determine which set of performance variables best discriminated between the performance from each 3-hour driving period. The resulting variables were then programmed into a micro-computer located on-board the test vehicle. Two weeks later, each subject returned to repeat the original 27-hour schedule. During this session, however, the performance of each subject was monitored on-line by the on-board computer. If their performance deteriorated below the criterion established from the data originally collected, the computer, through a D/A link, attempted to alert each driver. Results suggest that the multivariate criteria are able to discriminate between performance from the second 3-hour drive and that from the first at better than chance levels. The results are discussed in terms of an alertness indicator for highway vehicles.