Intelligent Vehicles > Collision Avoidance > Human Factors       Printer-friendly version

• Introduction
• Format and presentation of warning messages 
• Criteria of activation 
• Comparison studies
• Evaluation 
• References

INTRODUCTION

When evaluating the performance of different Collision Avoidance projects, it is important to look beyond the technological advances of the system and consider the affects on the driver. In attempting to improve the safety of travel as both technology and automobiles advance, we must ultimately decide if a Collision Avoidance System supports the driver's decision-making ability and driving tasks. To do so, many human factors must be taken into account to be sure that drivers can easily adapt to a new system or a new information display without jeopardizing other aspects of driving.

One of the most critical aspects of a collision warning and/or avoidance system is the system - driver interface. Laboratory and on-road tests have been conducted to evaluate the effect of different designs on driver reaction, and to compare these to driver behavior under normal (unassisted) driving conditions. Tests focus on two aspects of the design: the format and presentation of warnings and the thresholds for warning activation. Several measures of performance are used to qualify the different designs: braking time, minimum headway, average speed, and speed variability among others.

FORMAT AND PRESENTATION OF WARNINGS

These types of studies test the effectiveness and safety features of different in-vehicle devices used to communicate warning messages to drivers. One test consisted of simulating the approach to a stationary vehicle, with a warning message displayed using a four-second time-to-collision criterion. Five devices were tested:

• a fill bar type visual display (informative display condition) 
• abstract visual display (horizontal colored bars: green, amber, red) combined with a tone stimulus (single warning tone, 500 Hz, .86 sec.) 
• abstract visual display combined with a voice stimulus (Danger Ahead) 
• pictorial display (vehicle icon appearing at warning activation and replaced by a similar but larger one after .2 sec) combined with a tone stimulus 
• pictorial display combined with a voice stimulus 

Findings showed that the earliest braking response was obtained with the abstract/speech device. Most collisions occurred with the informative display device, while both abstract display devices resulted in less than half the number of collisions obtained in the control (unassisted) condition. A post-trial questionnaire indicated that drivers preferred the abstract/non-speech device.  

CRITERIA OF ACTIVATION

These experiments vary the parameters of the time-to-collision or worst-case scenario criteria. In one case (Hirst and Graham 1995), the study consisted of simulating the approach to a slower vehicle ahead. Three different warning activation times were employed (three, four, and five seconds time-to-collision), and subjects were instructed to brake at the last moment to avoid a collision. The main finding of this study was that subjects tend to brake earlier when presented with earlier warning.  

However it is not necessarily the case that an early warning results in earlier breaking action. A study by Janssen and Thomas (1997) found that a CAS which employs a three-second time-to-collision plus a one-second headway criterion performed worse (that is, lead to later braking action and more collisions) than a CAS which does not use the additional one-second headway. 
 

Findings indicated that the largest jump in annoyance ratings occurred between the conditions of four false alarms/hour and one false alarm/hour. It was also noticed that the voice stimulus device resulted in an average annoyance rating comparable to that of a tone device with a false alarm rate four times as large. Not much influence was seen between conditions #4 and #5, which suggests that drivers may accept warning systems even if nuisance alarm are not extremely rare. A simulation study (Farber, 1994) based on the Federal Highway Administration (FHWA) data and plausible braking and driver reaction time assumptions reported approximately 1100 false alarms per crash with an algorithm based on a minimum following distance criterion. If a rear-end crash occurs once every 50 years, this false alarm rate averages less than two false messages per month for each driver (the average western driver is involved in a rear-end collision about once every 25 to 30 years). 

COMPARISON STUDIES

Janssen and Nilsson (1991) tested simultaneously several different devices and criteria of activation. The CAS were evaluated in terms of their effect on four measures of driving performance: short headways with the preceding vehicle, average speed and speed variability of the CAS-equipped vehicle, and time spent driving in the opposite traffic lane. The experiment consisted of putting 8 groups of 7 subjects each in a simulator where they were told to drive on a standard two-lane road with four curves to the left and four to the right. Each group of subjects drove twice in the simulator, once without CAS and once with CAS. The eighth control group, drove twice without CAS. The simulator was programmed so that the CAS-equipped car would approach a leading vehicle moving at a specific speed, sufficiently often to assess the subject's behavior in dealing with that vehicle. In order to keep subjects from driving in the left lane all the time, obstacles were simulated on the left lane at irregular distances. 

Table 2 shows the characteristics of the CAS tested and the findings from the study. In summary, the authors concluded that: 

• All systems except one (#1, TTC + light) resulted in a reduction of short headways with respect to the control group. 

• All systems except one (#3, TTC + pedal) resulted in increases of average speed beyond that experienced by the control group. However, the standard errors are too large to show statistically significant differences. 

• The worst-case drivers experienced an average increase in speed variability with respect to the control group, while the TTC drivers either reduced their speed variability or maintained a variability similar to that of those in the control group. 

• All systems resulted in higher changes in acceleration and deceleration than the control group. 

• All systems resulted in drivers spending more time in the lane for opposing traffic. 

Findings from a subject questionnaire showed little correspondence between the results of the test and drivers preferences for the different systems. That is, those systems which resulted in the best driving performance did not rank, according to the qualitative questions, as the most useful or desirable. In addition, willingness to pay values varied from none to as much as $400 across the different systems.

Notes: 
1. TTC: four-second time-to collision  
2. WC: worst case, maximum deceleration rate 7 m/s2  
3. Cont: continuous display of a brake indication on the windshield  

Source: Janssen and Nilsson, 1991 

EVALUATION

Human factors issues associated with the design of crash avoidance systems include individual differences due to different population characteristics, cognitive and perceptual considerations, environment of vehicle use, and a number of other factors. By using experimentally-derived findings for driver response time to various warnings (audible or visual), a CAS can be evaluated in terms of its effectiveness in improving safety as well as the demands on the driver in possible pre-crash situations. Driver acceptance of a warning or automatic control is vital in the overall acceptance of a collision avoidance system. This acceptance is the key to behavior adaptation and the ultimate success of a CAS.

REFERENCES

Dingus, T., Jahns, S., Horowitz, A., and Knipling, R. Human Factors Design Issues for Crash Avoidance Systems. In: Human Factors in Intelligent Transportation Systems. Mahwah, N.J.:Lawrence Erlbaum Associates, 1998

Fukuhara, H. and K.Kurami. Essential Issues Involved in Radar-based Collision Warning-Avoidance Systems. In: IVHS America. Meeting (4th : 1994 : Atlanta, Ga.) The proceedings of the 1994 annual meeting of IVHS America. Vol. 2. Washington, D.C. : IVHS America, 1994.

Hirst, S. and R. Graham. The Format and Presentation of Collison Warnings. In: Ergonomics and safety of intelligent driver interfaces. Mahwah, N.J.: Lawrence Erlbaum Associates, 1997.

Huey, R. W., J.L. Harpster, and N.D. Lerner. In-Vehicle Crash Avoidance Warning Systems : human factors considerations Washington, D.C: National Highway Traffic Safety Administration ; 1997

Janssen, W. and L. Nilsson. An Experimental Evaluation of In-Vehicle Collision Avoidance Systems. In: International Symposium on Automotive Technology & Automation (24th: 1991 : Florence, Italy). 24th ISATA International Symposium on Automotive Technology and Automation, Florence, Italy, 20-24th May 1991. Croyden, England : Automotive Automation Limited, 1991.

Janssen, W. and H. Thomas. In-Vehicle Collision Advance Support Under Adverse Visibility Conditions. In: Ergonomics and safety of intelligent driver interfaces. Mahwah, N.J: Lawrence Erlbaum Associates, 1997.

Lerner N., D. Dekker, G. Steinberg and R. Huey. Inappropriate Alarm Rates and Driver Annoyance. Washington, D.C: National Highway Traffic Safety Administration ; 1996.

Schumacher R.W., R.D. Olney, R. Wragg, F.H. Landau, and G.R. Widman. Collision Warning System Technology. Traffic Technology International. 1996

Tijerina, L. Operational and Behavioral Issues in the Comprehensive Evaluation of Lane Change Crash Avoidance Systems. In: Transportation Human Factors. Mahwah, N.J.: Lawrence Erlbaum Associates, 1999.

Authors: Francois Granet and Rosella Picado. Last update: 02/27/01