Almost three years after completion, the National Highway Traffic Safety Administration (NHTSA) has released its final report on one of the most significant efforts of the 90's for the Office of Crash Avoidance Research:  the Automotive Collision Avoidance Systemprogram with Delphi-Delco, General Motors, and others.  A synopsis and highlights of the report are provided in this article.

The report details the results of the Automotive Collision Avoidance System (ACAS) program, which ran for three years from 1995-97.In this cost-shared project, NHTSA partnered with a consortium led by Delphi-Delco Electronic Systems, whose members included General Motors, Hughes Research Labs, ERIM International (now Veridian), the University of California - Davis, and STI.The report number is DOT HS 809-080.While electronic copies of such research results are normally released in advance of printing, in this case the report was delayed in all formats to give Delco time to file for patents.

For generalists, the 150-page report provides a fascinating look at the detail and rigor that must be included in collision avoidance system development to bring straightforward concepts (such as forward radar sensing) into a viable, effective vehicle system.For development engineers, the report is full of techniques, methodologies, and test results, although doubtless much of the "gold" was kept proprietary by the partners.

ACAS Phase One was birthed by the Technology Reinvestment Program sponsored by the US Defense Advanced Research Projects Agency (DARPA).According to the report's Executive Summary, the main objective of the Consortium was to provide a focused approach to accelerate the development of active collision avoidance systems.The program focused on implementing a "comprehensive" collision warning system, capable of detecting and warning the driver of hazardous situations to the forward, side, and rear of the vehicle.The system approach used long range radar and optical sensors for forward sensing, short range sensors for lane changes and backing up, and a lane detection system.Development was limited to providing warnings to the driver, not taking active control.

Far more than a proof of concept program, efforts in ACAS Phase One concentrated on cost reduction for manufacturing, development of immature technologies, and applying human factors engineering to ensure the system effectively interacts with the driver.The work also included crash analyses and development of demonstration/test vehicles.Human factors studies were conducted both on a simulator and in the test vehicles on a test track at the GM Technical Center.

"Remarkable Progress" Made

According to the report, "remarkable progress" was made in the areas of active sensors (radar, laser, and vision), algorithm/software development, and human factors.For instance, the linearity of the FMCW radar was improved by an order of magnitude, while the development unit cost was reduced by a factor of three (the production unit cost is projected to be reduced by a factor of five).A significant improvement in sensor reliability was also achieved.The collision warning processing algorithms matured, showing dramatic reductions in false alarms and missed warnings.

Figure 3.22: Performance Comparison (Lane Hunting/Wander Scenario) [click to enlarge]

Figure 3.27: Performance Comparison (In-Lane Vehicle on S-Curve Scenario) [click to enlarge]

A look at the table of contents shows that the report addresses problem definition, development of user requirements, development of near-term systems, approaches to cost reduction, studies of long-term advanced systems,and technical development of several subsystems: forward laser sensor, multi-beam planar antenna, low cost 24 GHz transceiver, lane sensing system, and a wide field-of-view heads-up display.The final sections of the report address development of warning concepts, simulator sensor models, driver-vehicle interface studies, and closed course testing.

While not the only ones in the industry tackling these problems, this team addressed many of the thorny issues confronting vehicle sensing systems -- false alarms caused by roadside signs or guardrails, curving road situations that would present a car in an adjacent lane as a hazard to a "dumb" radar, detection of a wide range of vehicle types (large trucks to motorcycles to tricycles), tuning of warnings to get the driver's attention without creating a startle effect -- and created a foundation upon which product rollouts are now occurring.

Electronic copies of the report are expected to be available very soon, in compact disk ROM format and downloadable from USDOT's Electronic Document Library (www.its.dot.gov).NHTSA's Jack Ference says that "plenty of copies are on the way" and can be ordered by contacting him.

[Top]

ACAS Report, Figure 1.1:  Conceptual Architecture for a Collision Warning System
[click to enlarge]
 

... contact Jack Ference of NHTSA at jference@nhtsa.dot.gov or Dr. Glenn Widmann at Delphi at glenn.widmann@delphiauto.com.

[Top]