The National Institute of Standards and Technology said today that two car collision warning systems it is testing have passed most, but not all, performance tests.
According to NIST, the systems passed most of the more than 30 tests conducted this fall but the systems had some warning system problems in detecting whether forward vehicles were in-lane or out-of-lane on curves or during lane changes. NIST also measured significant warning delays that resulted in test failures. Such problems are common in automotive crash warning systems that must operate in real-time, at highway speeds, and use multiple low-cost sensors to measure complex three-dimensional scenes, NIST said.
NIST said the Department of Transportation still plans to trot out 20 cars and 10 trucks with the warning systems to volunteers for further real-world testing in 2008. The DOT formed a partnership with the automobile industry called the Integrated Vehicle Based Safety Systems (IVBSS) initiative to hasten deployment of advanced warning systems.
The overarching goal of the IVBSS program is to combine existing safety and collision avoidance systems into an integrated system that can warn drivers of imminent crashes. A number of existing technologies are being utilized to address the three crash avoidance scenarios; these technologies will be further developed to work together to establish the integrated system:
· Rear-end Collision Avoidance: GPS, digital mapping, forward looking radar, and on-board cameras are used to warn drivers that they may be approaching the vehicle ahead of them at too great a speed to stop before crashing.
· Road Departure Collision Avoidance: vision based line tracking, and map-based road geometry, are used to warn drivers if they are about to drift off the road and crash into an obstacle, of if they are traveling too fast to for an upcoming curve.
· Lane Change/Merge Collision Avoidance: forward, rear and side-looking radar and vision-based cameras are used to determine if it is safe for the driver to change lanes or merge into a lane from an entrance ramp.
All systems use software that monitors vehicle speed, braking and throttle output, and provide messages to the driver if necessary, according to the DOT Web site.
To evaluate the performance of the crash warning systems, which generally use radar, researchers needed an accurate measurement tool based on entirely different principles. NIST researchers developed an independent measurement system (IMS) consisting of a camera and microphone in the cab to detect the driver warning, a suite of calibrated cameras to measure the distance to lane boundaries and laser scanners to measure the distance to obstacles forward and to the side of the vehicle.
According to DOT, of the 3.6 million rear-end, running off the road and lane change crashes that occur each year in the United States, 27,500 result in one or more fatalities--about three-quarters of the nation's yearly auto-related deaths. DOT estimates that widespread deployment of advanced integrated driver assistance systems may reduce such collisions by 48 %.
Layer 8 is Network World's daily home for the not-just-networking news.
Contact Layer 8
Layer 8 archive
The opinions expressed in this Weblog are those of the writer and may not represent the opinions of Network World.
|
|
This won't work at all in Mass.
The NIST's definition of a safe lane change is far different than that of a Massachusetts driver... If I had one of these installed on my daily commute, it would be blaring all the time. Like they say in NASCAR circles, it's not tailgating, it's drafting.
Personally I can't fathom
Personally I can't fathom relying on a car to tell me its OK to change lanes, rather I would like the car to tell me that it is NOT OK to change lanes, and I don't think this is a subtle difference. False positives are entirely acceptable, visual confirmation is easy and painless. False negatives are NOT acceptable and I fear that drivers would trust the system in this manner.
In practice, its great to make the system reliable, but it should be tuned for false positives. This is tricky because to be useful the driver has to have some level of trust that they don't just turn it off, but it should operate at that threshold, not at the opposite extreme that you rely on it blindly. Ideally you might even make a 100% reliable system and then statistically introduce false positives to keep the driver honest.
Audi and Mercedes already implement radar based 'adaptive cruise control'. From what I remember, Audi alerts you that you're coming up 'too fast' and pulses the brakes in an attempt to prompt the driver to action, but does not automatically brake for you. Mercedes will automatically brake for you and even bring the car to a halt, although I don't think you can drive 60mph at a brick wall and expect to stop, but the idea is to be useful in stop and go traffic.
Post new comment