A Transit Collision Avoidance Warning System (CAWS) Improved the Overall Driving Performance of Bus Operators Through System Alerts; However, Results Were Mixed Regarding Operator Satisfaction and Their Perception of Safety Benefits.

A Virginia DOT Evaluated Pilot Deployment.

Date Posted

Evaluation of a Transit Bus Collision Avoidance Warning System in Virginia

Summary Information

Even though bus operators are trained extensively, they can become distracted at times or fail to detect a vulnerable road user. In 2017, the Virginia Department of Rail and Public Transportation (DRPT) initiated a demonstration project to plan, implement, and evaluate a transit bus collision avoidance warning system (CAWS) on up to 50 buses. The CAWS included an advanced driver assistance system that utilized kinematic sensors and multiple external camera sensors to provide visual and/or audio alerts in various categories such as daytime pedestrian and bicyclist detection, warning for exceeding the speed limit, lane departure warning, and headway monitoring, and forward collision warning (FCW). The goal of the study was to evaluate the CAWS deployment at multiple transit agencies in Virginia in terms of system effectiveness and bus operator acceptance. The system was installed on 51 buses at nine transit agencies.


The CAWS logged various events, which were transmitted to a central database accessible to managers. When warnings were not displayed to the operator, the CAWS was referred to as operating in “stealth” mode, and when alerts were displayed to the operator, the CAWS operated in “live” mode. All equipped vehicles initially operated for at least two months in stealth mode, followed by at least six months in live mode. The study analyses focused on differences in event rates between these two modes on bus routes with high pedestrian activity in contexts including university towns, rural and suburban areas, and major metropolitan regions.

The study employed two types of data. (1) Quantitative telematics data obtained from CAWS logs, and (2) Qualitative data that consisted of operator feedback sessions during stealth and live modes which included 151 operator survey responses from transit agencies. In analyzing the quantitative data, the primary metric for evaluating the performance of the CAWS was the difference in mean events per 100 vehicle-kilometers traveled between stealth and live mode operation. The events in focus were collision warnings including FCWs and pedestrian collision warnings (PCWs), danger zones, dangerous driving behavior based on vehicle speed and longitudinal/lateral acceleration, and aggressive driving behavior.


  • Event rates were lower in live mode than in stealth mode by as much as 67.4 percent for one agency, which suggested that operators were generally responding to the alerts by allowing more space around pedestrians and reducing speeds and longitudinal/lateral accelerations that triggered dangerous and aggressive behavior events.
  • Changes in collision warning rates were mixed, with four agencies showing reductions (varying between 22.2 and 76.4 percent) and two agencies showing increases (varying between 0.2 and 10.8 percent).
  • The rate of danger zone events decreased in live mode (varying between 9.3 and 69.1 percent) with the exception of one agency where it increased by 12.9 percent.
  • Similar mixed results were observed with the number of dangerous behavior event rates where it decreased in live mode by up to 46.9 percent, except for one agency where it increased by 57.8 percent.
  • The operator survey revealed that while 80 percent of respondents indicated that the system was helpful, 76 percent of the survey respondents found the system distracting.
  • Seventy-five (75) percent of operator survey respondents indicated that they often or sometimes noticed false alarms and 73 percent stated that they took action to avoid vulnerable road users when the system did not give an alert or warning. 
Results Type