Experience with prototype testing on San Francisco freeways
This study examined driver behavior when using an advanced in-vehicle traveler information system designed to provide drivers with auditory warnings of slow moving traffic ahead on busy freeways in the San Francisco Bay Area. Field testing was conducted with 24 volunteer drivers who drove four vehicles equipped with a prototype system designed to receive real-time traffic information from external sources, assess vehicle speed differentials and position relative to approaching traffic queues, and provide drivers with auditory alerts such as "Slow Traffic Ahead" to prevent drivers from being surprised and causing hard braking events.
Research conducted prior to testing indicated that drivers who have 60-90 seconds of warning prior to negotiating traffic queues execute fewer numbers of hard braking events. In this investigation, researchers wanted to evaluate associated safety benefits; however, since crash events were relatively rare in these cases, and difficult to track without high market penetration and extended study periods, researchers decided to use driving behavior as surrogate measure to evaluate safety before and after the system was activated for each driver in each vehicle.
During this experiment four vehicles were equipped. Each vehicle was outfitted with a data acquisition system (DAS) to record both engineering and video data to characterize vehicle motions, driver behaviors, and the operation of the in-vehicle Networked Traveler Foresighted Driving alert system installed. Vehicle and driver data were collected seven days prior to system activation in each vehicle and seven days after activation. Project testing lasted a total of 17 weeks. In addition to analyzing vehicle and driver observations, participants were asked to fill out a short, web-based survey for each audible driving alert they received. As an incentive for participation, participants were paid $100 and reimbursed for fuel costs during the study period.
A total of 766 trips were recorded prior to activation (audible muted) and 782 trips were recorded after activation. 718 audible alerts were issued to drivers on more than 293 trips when the system was activated. A trip was defined as the time from when the vehicle ignition was turned on until the time when the vehicle ignition was turned off.
The presence of an audible alert system enhanced drivers' situational awareness and influenced the smoothness of driving profiles as drivers approached slower moving vehicles or vehicles that were completely stopped at end-of-queue traffic on the freeway. Vehicle and driver performance were evaluated using several metrics including: peak deceleration rates; average deceleration rates; deceleration due to braking; pre-braking deceleration; time before the start of braking; and the root mean square (RMS) error of speed metric which measured variability in speeds as drivers approached a queue The RMS Error of Speed metric supported the test hypothesis with the greatest statistical significance and confirmed that during the alert week drivers exhibited smoother driving profiles compared to the non-alert week. This metric was closely related to the standard deviation of speed metric used in previous traffic calming literature that shows for each 1 mph increase in the standard deviation of speed there is an 8.4 percent increase in the crash risk (Zheng, Ahn, and Monsere, 2010).
Overall, feedback from participants was positive. When the system worked as intended, 83 percent of the alerts received by drivers were rated as either good or neutral, and only 13 percent were rated as bad. When alerts were rated as bad, the typical reasons included wrong speed, late alert, or some indication that the traffic conditions were either expected or otherwise not worth reporting to the driver.