Automotive Collision Warning Effectiveness:
A Simulator Comparison of Text vs. Icons

Herbert Yoo, David Hunter, and Paul Green

To demonstrate the usefulness of an enhanced driving simulator capable of displaying traffic, 24 drivers participated in an experiment evaluating alternative formats for collision warnings that might appear on a head-up display. During a 15 minute drive (at 35 mi/hr), subjects passed 17 parked cars and 8 oncoming cars, and briefly followed 2 cars (that separately pulled out from the side of the road and then sped off). Subsequently, subjects encountered a car that pulled out from the roadside unexpectedly, requiring subjects to brake and/or swerve to avoid a collision. Each of the 3 warning conditions was responded to by 8 drivers. There were 3 collisions (out of a maximum of 8, 1 per driver) when an icon warning (showing the path to take) was presented, 5 collisions in response to a "swerve left" text warning, and 7 when no warning was presented. Additional data were collected for a second encounter (after another 13 minutes of driving); however, the crash rates were very low (the car pulling out was not a complete surprise), calling into question the value of exposing a subject to multiple scenarios for collision avoidance experiments.

Also explored were alternative measures of collision avoidance: subjective ratings of helpfulness, lateral clearance, impact speed, lane crossing position, throttle release distance, braking onset distance, and peak lateral acceleration. These measures, all ratio scale values, were thought to be more sensitive than collision counts to warning design differences. They were not. There were no significant differences in them due to warning format, suggesting that the number of crashes should be the primary performance measure.

Graphical Abstract (.pdf) | UMTRI-96-29 Full Report (.pdf)


Close This Window