In the late 1980s, the Washington DC capital beltway region experienced several severe accidents involving truck rollover accidents on curved entry/exit ramps. As a result, two major interchanges and one exit ramp were selected as test sites for deployment of a prototype automatic truck ramp rollover warning system (ATRWS).

• I-495W/I-95S in Springfield, Virginia.
• I-495W/Route 123N in McLean, Virginia.
• I-495E/I-95N in Beltsville, Maryland.

The systems were designed to alert truck drivers to slow down when maximum safe speeds were exceeded as they approached curved entry/exit ramps. Data from speed, weigh-in-motion (WIM), and height detectors were used to determine each trucks rollover threshold (RT) critical speed at the point of curvature for each ramp. If the software determined an approaching truck was at risk, then a fiber optic message sign was activated to warn drivers to slow down.



The impacts of the warning system were tested during a three year evaluation period (1993-1997).



IMPACT ON TRUCK SPEEDS



Two test periods were used to determine if the system was effective at reducing truck speeds at each site. The first test period occurred in May and November of 1994, and the second test period occurred in May, July, and October of 1996 (16 months after installation). Speed detectors were located upstream and downstream of the warning sign to measure changes in truck speeds. The evaluation results indicated 500 trucks activated the warning sign and reduced their overall average speed by 8.3 miles per hour (mi/h). The 252 trucks that did not activate the warning sign reduced their overall average speed by 6.5 mi/h.



IMPACT ON ROLLOVER ACCIDENTS



To evaluate if the rollover warning systems were effective at reducing the number of rollover accidents at each test site, crash data were collected before and after the systems were deployed. Virginia provided before (historical) crash data (1986-1989) for the Springfield and McLean sites, and Maryland provided before crash data (1985-1990) for the Beltsville site. To estimate the potential impact of the system on rollover crashes at each site, a Poisson Probability Distribution model was used. Based on an analysis of the before data, the model estimated that each Virginia test site would experience 0.5 rollover accidents per year if the system was not deployed, and the Maryland site would experience 1.2 rollover accidents per year if the system was not deployed.



Based on the fact that no rollover accidents occurred at any of the 3 test sites during the 3 year evaluation period, the evaluation results indicated there was a 0 percent probability of experiencing 1 future rollover accident in 1 year at each of the sites tested.