Methodology

Four test cases were defined to represent scenarios common in fatal pedestrian-vehicle crashes:

• Test Case 1: a vehicle is traveling straight, on a straight road, and a pedestrian/cyclist makes a perpendicular crossing.
• Test Case 2: a vehicle attempts a right turn at an intersection while a pedestrian/cyclist attempts a straight path roadway crossing.
• Test Case 3: a vehicle attempts a left turn at an intersection while a pedestrian/cyclist attempts a straight path roadway crossing.
• Test Case 4: a vehicle is traveling straight, on a straight road, and a pedestrian/cyclist is traveling straight along the roadway.

Three speeds, 10 mph, 15 mph, and 20 mph, were used to vary approaching vehicle speeds in test cases and ensure safe test execution. Only test cases 1 and 4 were conducted for the camera-based aftermarket system because preliminary trials in test case 2 and 3 yielded unreliable pedestrian alerts. Trials were performed at the signalized intersection and marked mid-block crossing locations at the test bed. Test case 4 was used to assess the camera-and-radar-based OEM system as the limitations of the system (noted in the vehicle manual) prevented the assessment of test cases 1 through 3. Camera-and-radar-based system trials were conducted at two marked intersection crossings at the test bed. The smartphone-based application was tested for test cases 1 through 3 at four marked, signalized crosswalks at the test bed with five trials in each crossing direction.

Various measures were considered in assessing the three systems, including accuracy and reliability of detection and alerts, as well as other factors such as user interface and notification quality, response selection, user access to technology, readiness, known non-functional situations, ease of use, and interpretability of warning. Performance measures used in this testing were tailored to the specific systems being assessed, and included:

• Percent of trials where alert was issued (cautionary and emergency alerts)
• Average distance from crosswalk when alert was issued (cautionary and emergency alerts)
• Average distance from crosswalk where vehicle came to a stop
• Time to Collision (TTC), calculated using the measured distance from crosswalk divided by the assigned vehicle speed
• Percent of trials with supplemental automated braking applied
• Percent of trials with full automated braking applied
• Percent of trials with the correct pedestrian guidance message presented

Findings

• Test results indicated a potential safety benefit offered by the two vehicle-based systems when clear and direct sightlines were available. On average, the camera-based system issued cautionary visual alerts when the vehicle was 38 feet away from the crosswalk, and the emergency alerts allowed sufficient time for braking prior to the pedestrian target. The camera-and radar-based system issued alerts in 97 and 63 percent of pedestrian and bicyclist trials, respectively.
• Testing indicated that the forward-looking camera-based safety device used in this experiment detected pedestrians and produced alerts with adequate timing and clarity to prevent collisions. Although the system failed to produce alerts in some trials, when detection was achieved, it provided advance notice to increase the time the driver could react to avoid a crash.
• The camera-based system’s ability to detect and warn drivers of at-risk pedestrians could be useful for reducing crashes in a variety of scenarios, such as when the driver is distracted from the roadway, and in crowded urban settings. This system is most beneficial for reducing daytime crashes on roads where speeds are 31 mph or lower.
• The smartphone-based system is not reliant on external sensors, and its functionality was found to be independent of light conditions, roadway grade or curvature, and pedestrian density.