Future in-vehicle crash avoidance warning systems will inevitably deliver inappropriate alarms from time to time, caused for example, by situations where algorithms have correctly identified an object but pose no threat or danger to the driver. The current state of knowledge does not permit an estimate of how many inappropriate alarms users find unacceptable, and how that rate may vary with factors like the type of signal generated by the system (i.e., tone versus voice), or extended experience with the warning system itself. The purpose of this study is a direct comparison of drivers' subjective annoyance towards inappropriate alarms as a function of rate of occurrence and the type of signal generated in naturalistic, on-road driving conditions. Test equipment to generate and present signals, and to collect driver response was installed in fifteen participants' personal vehicles for a nine week period. Signals were presented at random times while the participants engaged in their normal, daily driving routines. In order to simulate future operating conditions where actual alarm warnings will require the driver's attention and reaction, "appropriate" alarms to which the driver had to make a simple motor response, and "inappropriate" alarms to which the driver did not have to make any response, were presented. Inappropriate tonal alarms were presented at four different frequencies of occurrence, including averages of four per hour, one per hour, one per four hours, and one per eight hours of driving time. In addition, a voice warning condition was included, at a rate averaging one per hour. Participants made daily and weekly ratings of the degree of annoyance that resulted from the nuisance alarm schedule. The 4/hour tone and the 1/hour voice were significantly more annoying, and less acceptable, than the other conditions. Participants showed a wide range of annoyance sensitivity, but the two most annoying conditions appear to be unacceptable, while the less frequent rates do appear potentially reasonable for functional systems. (Author/publisher)
Abstract