Advanced Driver Assistance System Implementation on Buses in Florida Led to an Average 34 Percent Reduction in Detected Conflicts.

Driver Assist System Implemented on Ten Buses in Florida Leads to Reduced Vehicle-to-Pedestrian and Vehicle-to-Vehicle Conflicts with Improved Safety Impacts.

Date Posted
10/21/2022
Identifier
2022-B01686
TwitterLinkedInFacebook

University of Florida (UF) Testbed Initiative – Alternative Transportation Safety Systems

Summary Information

Advanced Driver Assistance Systems (ADAS) have the potential to decrease crash risks or conflicts associated with public transit operations, which could lead to a safer, more efficient, and economical service. This study evaluated the safety benefits of a vision-based ADAS, procured and installed on ten Regional Transit System (RTS) buses in Gainesville, Florida, in January 2019. The system included camera sensors, two exterior sensor housings, two interior windshield-mounted vision sensors, and three driver alert displays that communicated with the driver by providing visual and audible alerts. A before-and-after analysis was conducted over a period of one year of operation focusing on the pedestrian detection (PD), pedestrian collision warning (PCW), urban forward collision warning (UFCW), forward collision warning (FCW), headway warning (HWW), and aggressive braking (AggBrk) alerts. In addition, 18 transit operators were interviewed through a focus group study to obtain feedback. Lastly, the study developed a benefit-cost analysis tool using crash history along with surrogate safety assessment parameters to predict the rate of return of ADAS investment.

METHODOLOGY

The ADAS operated in stealth mode between January 4 and March 4, 2019, during which drivers did not receive alerts ("before" period). The system then operated in active mode between March 5, 2019, and February 29, 2020, during which drivers received alerts ("after" period). The daily logs of all alert data from seven of the ten buses were included in the before-and-after analysis (three of the buses were excluded from analysis due to low/no data). The data analysis was comprised of an aggregate analysis including conflicts from the buses, and a route-based analysis. For the aggregate analysis, all the vehicles’ combined alerts were analyzed to understand the hourly, daily, and monthly distribution of alerts and to compare the differences between the before and after periods. For the route-based analysis, a visual methodology was devised to determine each bus's route on each day using the alerts’ latitude and longitude readings. Thirty-two routes of seven buses were extracted using this visual methodology.

For the transit operator interviews, five focus group sessions of four bus drivers each were held to elicit responses from bus drivers who had experienced the ADAS. For benefit-cost analysis, input parameters included transit data, safety information, and monetary values of investments and crashes. Using the input information and predefined values for converting surrogate safety measures to a predicted number of crashes, the net benefit and benefit-to-cost (B/C) ratio were computed.

FINDINGS

  • Aggregate analysis revealed a 34.17 percent average of reduction in the number of alerts for the six alert types, which translated to an average conflict modification factor of 65.83 percent. The three largest reductions were for HWW, AggBrk, and PCW at 48.30 percent, 47.60 percent, and 33.40 percent reduction, respectively. These results indicated that with the introduction of ADAS, the number of conflicts was reduced.
  • Route-based analysis showed an average reduction of 19.95 percent in the number of alerts.
  • From the bus driver focus groups, it was concluded that about 75 percent of the drivers found the system useful, especially on the University of Florida campus where the students increased the congestion levels. Seventy-six percent (13 out of 17) of the respondents said that they felt the ADAS improved safety. Despite the overall usefulness, many drivers reported too many alerts and alarms with no reason (false positives), to the point that it became distracting and reduced drivers’ trust in the system.
  • Benefit-cost analysis revealed a lower bound B/C ratio of 1.5 and an upper bound B/C ratio of 4.38 for Gainesville, FL for the ADAS implementation.

 

Results Type
Deployment Locations