Design, Simulation, & Testing of a Pedestrian Alert System (PAS)

Charles Rodgers, Darrell Geenlee, Robert Nelsen, George Likakis and Richard Blomberg

Abstract:	The Federal Highway Administration (FHWA) is seeking to develop a system to alert automotive drivers to an impending collision with a pedestrian. Under a Small Business Innovative Research (SBIR) contract to FHWA, OPTIMUS Corporation and its subcontractor Dunlap and Associates have designed, simulated, and field tested a prototype PAS that relies on DGPS techniques to accurately determine relative position and velocity measurements, which allows the prediction of collision probabilities, and activation of an appropriate driver alert. The key requirements of the system include providing accurate warnings with a low false alarm rate at least three seconds before the collision, all weather functionality, and the capability of detecting pedestrians hidden by visual screens. It minimizes false alarms through a variable "warning zone" that depends on vehicle and pedestrian motion. It also is capable of alerting the pedestrian in danger. The PAS design is compatible with Intelligent Transportation System (ITS) and telematics equipment. Most vehicles sold in the future will have a GPS, a processor, and eventually an ITS wireless data link in their telematics, which will minimize the PAS equipage costs. The PAS has both pedestrian and vehicle modules, each containing a low power, GPS receiver capable of weak signal tracking, and a low cost/power, radio transceiver. The pedestrian module broadcasts its position and velocity information to all vehicles in range with no personally identifying information. The vehicle module receives the information from the pedestrian and determines the relative distance and velocity to the pedestrian. The vehicle unit calculates a "warning zone" based on its velocity vector, the state of turn signals and transmission, and the velocity of the individual pedestrian. If the pedestrian is predicted to be in the warning zone and the time-to-collision is within a critical time (on the order of 6 seconds), the driver receives an alert. The alert provides information on both the urgency and direction of the threat. OPTIMUS performed extensive DGPS simulations on scenarios based on standard pedestrian crash types (e.g. child dart out). A flexible model of buildings and other potential signal obstructions were included to simulate the effects of urban canyons and other non-ideal environments. Each simulation was run on the chosen scenario repeatedly over 24 hours of satellite motion, and statistics were calculated. Simulation results include expected differential position and velocity accuracy for different implementations of DGPS, and with differing numbers and heights of buildings. One of the most challenging aspects of the system design was developing alert algorithms that accounted for realistic traffic conditions and GPS uncertainty, while minimizing false alarms, and providing timely alerts. This algorithm was developed and integrated into the GPS simulation. The calculated alert state at each second of simulation was then collected and analyzed while varying the various GPS error types and levels. The results indicated that the system could deliver high integrity alerts with a good false alarm rate. The most challenging aspects of the hardware design were the radio link, driver alert implementation, and the engineering of the small pedestrian module. The radio link must allow unlicensed transmitters but provide high integrity communication over distances sufficient for GPS acquisition and tracking before the point of closest approach, while maximizing battery life. The driver alert must be interpretable in an instant with little training and it must not cause a dangerous reaction from the driver. Candidate system components were tested individually, and the best performers were chosen for the final design. OPTIMUS tested several weak signal GPS receivers in various environments including under heavy foliage and in canyons. Tracking characteristics were compared. Amplitude Shift Key (ASK), On-Off-Keying (OOK), FM, and frequency hopping FM radio links were tested in benign and noisy environments to assess each one’s Bit Error Rate (BER). Range versus BER was determined and plotted. In addition, an FM link was simulated using a commercial communications design package. The chosen RF transceiver has sufficient range to provide more than eight seconds of GPS tracking before the point of closest approach. The goal of the pedestrian module design was a watch sized device that could be worn by a person that would provide more than 24 hours of protection between recharges. Meeting the power consumption goal required developing elaborate conservation techniques including millisecond control of the operation of the RF transceivers and the GPS receiver. The real-time application software for the pedestrian module is embedded in the GPS receiver, while the vehicle module uses an external single board computer to host its real-time software. The paper will present the concept of operations, key design tradeoffs, final system design, GPS and alert simulation results, and component test results.
Published in:	Proceedings of the 17th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2004) September 21 - 24, 2004 Long Beach Convention Center Long Beach, CA
Pages:	2447 - 2457
Cite this article:	Rodgers, Charles, Geenlee, Darrell, Nelsen, Robert, Likakis, George, Blomberg, Richard, "Design, Simulation, & Testing of a Pedestrian Alert System (PAS)," Proceedings of the 17th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2004), Long Beach, CA, September 2004, pp. 2447-2457.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In