Safety Afoot

May 1, 2008 By: Charles Rodgers, Richard Blomberg GPS World

Field Testing of a Pedestrian Alert System

The driver of the future may enter an automated roadway and turn the controls over to system automation. Under this scenario, a technologically based pedestrian detection and tracking system will be essential if pedestrians are to share the roadspace. Some means other than the visual acquisition of a pedestrian by a driver will be required to notify the entire system of the presence of the pedestrian.

Even with humans at the controls, some alert system is needed to reduce the alarming casualties: more than 13 pedestrians killed and 167 injured per day in the United States during 2006, according to National Highway Traffic Safety Administration (NHTSA) statistics. Over the last decade these numbers have been relatively unchanged.

NHTSA found that motorists' failure to detect pedestrians was often the primary crash cause.

Many automobile manufacturers, also concerned about pedestrian collisions, have announced prototype detection systems as well as material and design changes to vehicles to minimize pedestrian injuries in case of a collision. However, most detection systems rely on infrared or radar technology to detect pedestrians. These can be helpful but they do not discriminate at-risk pedestrians from other people and objects in the area, so they are subject to false alerts. In addition, these systems cannot detect a pedestrian behind a visual screen such as an ice-cream truck or parked cars. The pedestrian alert system (PAS) described here can do that as well as prioritize pedestrians (peds) by risk.

Concept of Operations

The key technical requirements of the system include:

1. All-weather, non-line-of-sight detection of peds
2. Classification of peds according to risk of collision
3. An effective driver alert, more than three seconds and less than eight seconds before crash
4. Minimization of false alerts and missed detections.

The opening figure illustrates the PAS concept. A vehicle (1) is equipped with a GPS/GNSS receiver, processor, and a short-range (less than a quarter-mile) radio transceiver, while both peds A and B are equipped with a compact module that contains a GPS/GNSS receiver and a radio transceiver. As the vehicle travels along the road, the radio transmits a wake-up signal to any equipped peds in the area. Upon receiving the wake-up signal, the ped's GNSS receiver begins to acquire satellites in a hot-start mode.

When a position and velocity solution is available, it is transmitted to the vehicle, where it is determined whether an alert to the driver and the specific ped is appropriate. The processor in the vehicle uses differential GPS/GNSS algorithms to determine accurate differential position and velocity vectors, and calculates the expected relative position vector at each of the next eight seconds. If the ped end of the vector is calculated to be within the warning zone (represented in the figure by a rectangle), determined based on the vehicle and ped velocities, a warning is given to the driver and transmitted to the at-risk ped. If the ped is determined to be out of the warning zone, they are judged to be a low priority and may be tracked at a lower update rate. Since peds can be near the road but not in danger, warnings are not issued based on proximity. A warning is only issued if the ped is actively moving on a collision course with the vehicle (or vice versa).

As shown in the figure, the system works regardless of the visual screen formed by the shrubbery. At the instant depicted in the figure, ped B is outside the wake-up signal range so his receiver and transmitter remain asleep to conserve battery power.