Lesson

Incorporate proven technologies and false alarm reduction strategies in the design of future Automotive Collision Avoidance Systems (ACAS).

Experience from a Field Operational Test (FOT) in Michigan to evaluate an Automotive Rear-End Collision Avoidance System.


April 2006
Michigan,United States


Background (Show)

Lesson Learned

This lesson is drawn from the experiences of the Field Operation Test (FOT) of an Automotive Collision Avoidance System (ACAS). Overall, the ACAS FOT program was successful in building a production-intent rear-end crash avoidance system on-board a passenger vehicle. This system integrated state-of-the-art technologies that performed forward collision warning (FCW) and adaptive cruise control (ACC) functions. In addition, this program produced a reliable small fleet of ACAS-equipped vehicles that were used by lay people in an FOT as their own personal cars to experience ACAS functions under different naturalistic driving conditions. Given the scope of the program in terms of its duration and size of the vehicle fleet, the FOT was also successful in building a knowledge base about driver performance with and without ACAS assistance from 97 percent of the distance traveled during the FOT and about drivers’ opinions of the ACAS. Based on FOT and system characterization test data, the independent evaluation was able to delineate the strengths and limitations of ACAS capability, gauge driver acceptance, and assess its safety impact. The FOT provided a first opportunity to obtain real world feedback from drivers about their tolerance of nuisance and false crash-imminent alerts. Moreover, both positive and negative safety consequences of ACAS use were highlighted. Indicators of positive safety impact outweighed those of negative safety impact. The independent evaluation was partially successful in projecting potential safety benefits of ACAS by combining FOT data with national crash statistics, which were constrained by short-term use of ACAS by relatively few subjects. Below are general comments reflecting on past and future FOTs of crash avoidance systems, based on this independent evaluation.
  • Consider state-of-the-art system design issues and technologies for Automotive Collision Avoidance Systems (ACAS). The FCW function of ACAS incorporates state-of-the-art sensor technologies for short-term deployment plans (1 – 2 years). However, improved signal processing and threat assessment algorithms would enhance FCW alert efficacy by recognizing slower lead vehicles transitioning from the path of the host vehicle to out of its path. This event generated numerous unnecessary crash-imminent alerts during the FOT, and even forced the ACC to automatically brake in response to lead vehicles exiting the freeway. Stationary out-of-path targets were mostly the source of false crash-imminent alerts. The GM Consortium identified some remedies that seemed to be worthy of consideration in dealing with this particular problem, including the disregard of the closest in-path stationary (CIPS) target flag by the target selection algorithm. The remedy is for the threat assessment algorithm to rely completely on the closest in-path moving (CIPV) target flag that accounts only for moving vehicles and for stopped vehicles tracked by the radar to be moving prior to stopping. This approach would increase system credibility and driver acceptance since false alarms to these stationary (never before seen moving) objects would be removed. The examination of video episodes revealed a few cases where CIPS-tagged vehicles triggered the crash-imminent alerts, mainly at intersections. Thus, a concern is raised regarding the elimination of the CIPS flag from the threat assessment algorithm.
    The analysis of crash-imminent alerts also showed that increasing the threshold operating speed of FCW over 25 mph would not make any significant impact on false and nuisance alerts (> 50% reduction). To boost driver acceptance of FCW at the expense of some limited safety benefits, it is recognized that a trade-off must be made between alert rates and the operating envelope and sensitivity of FCW. The ACAS incorporated many subsystems to identify the path of the host vehicle, and track and select targets at long ranges in the path of the host vehicle. One of these subsystems is GPS/GIS mapping to help identify the path of the host vehicle and make in-path target selection. It appears that this feature had little impact on crash-imminent alerts as was evident from the system characterization test that was conducted in the Boston metropolitan area. The map information was not available there and the alert rate did not seem to differ from the rates observed in Michigan by FOT subjects with available map data. Given the cost of such a feature, the ACAS could perform without it unless, of course, this feature is also a part of a navigation device or a curve speed warning system. Moreover, it is recommended that human factors tests be conducted to obtain user feedback on the usability of some of the HUD icons presented to FOT subjects by the ACAS. This recommendation is based on qualitative comments made by FOT subjects during debriefings and focus group meetings. It should be noted that only the cautionary and crash-imminent alert icons of FCW were tested prior to building the pilot vehicle for the FOT. Survey and subjective data from FOT subjects and system characterization test data suggest that even better acceptance of ACC would be achieved with improved automatic acceleration and deceleration characteristics. The results of the independent evaluation suggest marginal acceptance of FCW and better acceptance of ACC as well as some positive safety indicators that warrant deployment at least at low-level market penetration.
  • Conduct additional research to reduce false alarm rates. Additional research may be necessary to reduce the rates of false and nuisance alerts of FCW and to enhance the timing of crash-imminent alerts for mid-term deployment plans (2 – 5 years). Proceeding with further FCW enhancement activities may depend on successful results (driver satisfaction, units sold, and positive safety impact) from short-term deployment and good market penetration levels. The recognition of the driver state would improve FCW alert timing, ranging from low complexity to identify the location of driver face (facing forward or sideways), medium complexity to track the eyes of the driver, to high complexity to measure the cognitive load of the driver. This research could build on current efforts undertaken in the SAVE-IT program (Witt et al., April 2004). Another FCW improvement might be achieved with the use of digital image processing of the forward scene to discern the objects that the radar is tracking. This might reduce the rates of crash-imminent alerts due to stationary out-of-path targets. Vehicle to vehicle communication is suggested to improve the forward-looking sensing capability of FCW for long-term deployment plans (> 5 years). This research would build upon prior work in vehicle safety communications (Crash Avoidance Metrics Partnership, May 2004). This enhancement would call upon lead vehicles to transmit information about their state to following vehicles, given wider deployment of FCW in the vehicle fleet. The transmission of relevant information about the lead vehicle such as its dynamic state (stopped in traffic, moving at constant speed, decelerating, or accelerating), brake initiation, and value of its acceleration/deceleration might improve the timing of crash-imminent alerts, thus reducing the rates of "too late" alerts (increasing crash prevention potential) as well as "too early" alerts (decreasing nuisance alert rate). It should be noted that this current ACAS estimates the value of lead vehicle acceleration/deceleration in support of the timing algorithm. Proceeding with such system improvement activity might depend on significant market penetration rates of FCW in the vehicle fleet during the next 5 to 10 years.

The experiences from this FOT demonstrate that to develop and implement effective Automotive Collision Avoidance Systems, state-of-the-art technology and research issues should be carefully considered. For example, improved signal processing and threat assessment algorithms would enhance FCW alert efficacy by recognizing slower lead vehicles transitioning from the path of the host vehicle to out of its path. In addition, further research may be necessary to reduce the rates of false and nuisance alerts of FCW and to enhance the timing of crash-imminent alerts for mid-term deployment plans (2 – 5 years). Helping to solve these issues with Automotive Collision Avoidance Systems will presumably help to increase driver satisfaction, and increase the safety impact of these systems.


Lesson Comments

No comments posted to date

Comment on this Lesson

To comment on this lesson, fill in the information below and click on submit. An asterisk (*) indicates a required field. Your name and email address, if provided, will not be posted, but are to contact you, if needed to clarify your comments.



Source

Evaluation of an Automotive Rear-End Collision Avoidance System

Author: Najm, Wassim G., et al.

Published By: Prepared by Volpe for the U.S. Department of Transportation

Source Date: April 2006

EDL Number: 14303

URL: http://ntl.bts.gov/lib/jpodocs/repts_te/14303.htm

Lesson Contacts

Lesson Analyst:

Brian Philips
Noblis
202-863-3659
brian.philips@noblis.org


Rating

Average User Rating

0 ( ratings)

Rate this Lesson

(click stars to rate)



Notes

Lesson of the Month for September, 2008 !


Lesson ID: 2007-00328