Research Needs

Research is needed to determine whether signal faces mounted at the roadside will be effective in the roundabout context.

Research is needed to determine the minimum number of signal faces necessary to achieve high driver compliance.

Guidelines for computing the optimal location of the stop bar are needed.

U.S. Access Board has identified signalization as the only proven means of creating crossable gaps in traffic and providing audible cues to pedestrians of the availability of gaps. Although there is extensive experience with traffic signals both in the U.S. and abroad, experience with traffic signals at roundabout crosswalks is limited, and experience with signals to provide access at roundabout crossings to persons with visual or other disabilities is non-existent. Therefore, although it is reasonable to expect that accessible traffic signals will be effective at multilane roundabouts where access is otherwise problematic, the lack of implementation experience suggests the need for studies to document potential pitfalls and best practices.

U.K. experience suggests that drivers will not confuse the pedestrian crossing signals with the yield at entry control, but this needs to be verified with U.S. drivers.

It was suggested above that a signal face that is blank (off) until activated by a pedestrian would be more effective that a signal that flashes amber until activated. This suggestion was based on the assumption that drivers would learn to expect the flashing amber, and thus be less likely to detect when a pedestrian call had caused the light to become steady. The MUTCD recommends a flash rate of about 1 Hz (about 0.5 s on and 0.5 s off). The off cycle distinguishes flashing amber from a solid. It would probably take well in excess of 0.5 s for a motorist to detect that the off cycle was not coming, particularly if the motorist was not expecting a phase change. At a minimum, the onset of the flashing amber would reduce motorist recognition time by approximately 0.5 s. The reduction in reaction time might be much greater for motorists who are not anticipating a phase change. These considerations suggest an evaluation of the effectiveness of blanking out the signal when not in use versus flashing amber when there is no pedestrian call.

The precedent for a blank pedestrian crossing signal is the High Intensity Pedestrian Activated crossWalK (HAWK) signal that is currently in use in Tucson Arizona (33). This crossing control, which has approval from FHWA as an experimental device, consists of two red lenses over an amber lens. When a pedestrian presses the call button, the amber flashes for three seconds, then turns solid for four seconds. The solid amber is followed by two solid red lenses that are illuminated for the duration of the walk phase on the pedestrian signal head. At the beginning of the flashing do not start phase on the pedestrian signal head, the red lenses on the traffic signal begin alternating. Depending on the length of the walk phase, the HAWK signal may, or may not be appropriate for roundabout accessibility. Because the flashing red phase is a stop control—drivers may continue after coming to a complete stop. Under stop control, vehicles may proceed when it is safe to do so. Blind pedestrians often wait and listen for assurance that vehicles have stopped for them. Drivers may interpret this delay as a decision by the pedestrian not to cross, and proceed just as the pedestrian steps into their path. An APS indication of the crossing-interval might not be appropriate with a stop control, because vehicles will still be permitted to proceed after stopping. Because roundabout crossings are short, the flashing red phase may not be necessary with HAWK type signals at roundabouts. Evaluation of the HAWK crossing in Tucson, which is not at a roundabout, showed that driver yielding increased from about 34 percent with an uncontrolled crossing to between 93 and 97 percent with the HAWK, and that pedestrians were less likely to run or hesitate during their crossings after the HAWK was installed (34).