One of the less endearing things about computers is their readiness to tell you that you are doing something wrong. The computers are about to notch this up, and, like it or not, it could be a good thing. An annoying computer is preferable to being taken out by an 18-wheeler as you come off an Interstate ramp.
Researchers at MIT and Nissan are developing a car computer system that can predict when you are about to do something stupid and tell you so. Observing your behavior, the car knows if you are going to turn, change lanes, brake, or pass another car in the next few seconds. If this is not a good idea, based on other input, it will tell you so. Appropriate warning or override systems will be activated. At the same time, another group of researchers, at Oak Ridge, is studying automotive information overload. It wants to know how drivers, especially older ones, react to all of the things technology can throw at them through cellular phones, electronic mail, collision warning and navigation systems, head up display of day and night road imagery, adaptive cruise control, and lane trackers.
Tests using a driving simulator show that the MIT-Nissan warning system is 95% accurate at predicting drivers' moves 12 seconds in advance. It is based on the observed fact that simple driving behaviors can be broken down into long chains of simpler sub-actions, including preparatory action. If they think about it, passengers could probably predict what a driver will do next, and the computer, which does think about it, does just that. The smart Nissan uses a computer and sensors on the steering wheel, accelerator, and brake to monitor a person's driving pattern. A brief training session in which the driver performs certain maneuvers lets the system calculate the probability of particular actions occurring in two-second time segments.
A statistical technique similar to the one used in speech recognition software works out what's probable. The driver's behavior is continually monitored for patterns. When a pattern is recognized, the system predicts what the driver will do next. It continually updates the probabilities, so it gets more accurate the longer it monitors a particular driver. They haven't tried this yet, but the researchers think, based on some eye movement research they've done, that they can get all of the information they need by just tracking the driver's eyes. This would give an even earlier warning system.
The Oak Ridge team will study information overload's effect on drivers 18 to 75, driving a specially equipped 1999 Dodge Intrepid with all of the available information overloaders. Will the new information startle drivers? Physiological monitors will measure what's going on with the driver, providing information about a driver's reactions. Sensors on the wheels, steering wheel, and global positioning system will tell what the car is doing while video cameras will see what the driver is doing, and what's ahead and behind the car.
"Reprinted with permission of Technical Insights/John Wiley & Sons, Inc., New York, NY, 212-850-8600. http://www.wiley.com/technical_insights"
Details: Alex Paul Pentland, Media Lab,
Massachusetts Institute of Technology,
20 Ames St., Cambridge, MA 02139.Phone: 617-253-0648.
E-mail: sandy@media.mit.edu.Philip F. Spelt, Senior Research Scientist,
Oak Ridge National Laboratory,
PO Box 2008, MS 6364,
Oak Ridge, TN 37831-6364.Phone: 423-574-7472.
Fax: 423-241-0381.
E-mail: speltpf@ornl.gov.
'Microelectronics Technology Alert' tracks significant developments in electronics, optoelectronics, data storage and transfer, focusing on the manufacture of devices using techniques such as vapor deposition and ultralithography. Advances in semiconducting and superconducting materials, capacitors, light emitters and other materials are also monitored. Major changes in industry and novel designs for electronic components and systems are covered.
TECHNICAL INSIGHTS ALERT
Phone: 201-568-4744Contact: Heide.Roth@wiley.com
Use of this information is determined by license agreement; any unauthorized use is prohibited.
Copyright 1999, John Wiley & Sons, Inc.
ISSN 1084-4546