Untitled Document

The milk-chocolate brown 1989 Chevy Celebrity station wagon doesn't look like much more than an over-the-hill family car, with its boxy lines and dusty patina.

But to University of Arizona researchers, the so-called VISTA car represents the not-so-distant future of motoring: cars that drive themselves.

A team of UA engineering professors and students have built a system that allows the 10-year-old car to guide itself along a road. The car steers along a mapped road, adjusting its speed to safely stay in-lane and at a safe distance from other vehicles.

How it works: Smart car

* A radar unit mounted on the station wagon's front bumper senses objects ahead and sends that information to the central computer.

* A camera mounted inside the top center of the windshield scans the road for visual cues to the car's alignment within a lane, such as lane stripes. Additional sensors in the front and rear wheels help gauge turning angle and speed, respectively.

* Based on pre-loaded road data and information sent from the radar, camera and sensors, the computer constantly sends new commands to small motors that control the car's throttle, steering and brakes.

The state-sponsored research project, dubbed VISTA for Vehicles with Intelligent Systems for Transport Automation, aims to develop an ``affordable, intelligent vehicle'' that can be on the road within the next five to 10 years.

The project's first target: the two-hour commute between Tucson and Phoenix on Interstate 10.

Researchers at the UA's Advanced Transportation and Logistics Algorithms Systems (ATLAS) Research Center are designing a car that will automatically drive between the two cities using planned ``Intelligent Express Lanes'' on Interstate 10.

The Arizona Department of Transportation is sponsoring the project, with a $100,000 grant awarded by the Legislature last year.

The effort is part of a broader nationwide effort to develop so-called intelligent transportation systems, which also include road systems that relay traffic information to traffic controllers and motorists.

Steve Owen, an ADOT research project manager, said vehicle guidance systems could help make better use of available roadways and in some cases delay costly freeway upgrades.

Some elements of the guidance systems already are on the market, Owen said, including an ``adaptive cruise control'' systems no available on some Mercedes models that automatically adjust speed when a car encounters other vehicles.

Using a car borrowed from ADOT and commonly available hardware, the VISTA team led by UA Systems and Industrial Engineering professors Pitu Mirchandani and Fei-Yue Wang, with the help of several students, installed various systems that allow the VISTA car to drive itself. An Arizona State University engineering team installed the control motors on the car's throttle, brakes and steering.

Using pre-programmed road data and feedback from a radar mounted on the front bumper, a windshield-mounted camera and wheel sensors, the VISTA system controls throttle, braking and steering by means of small motors (see accompanying diagram).

``Everything's off the shelf,'' said Mirchandani, director of the ATLAS Center. ``The computer (a 200-megahertz Pentium desktop) is like you'll find on your desk.''

But building the brains of the system - a collection of controllers, computer code and mapping data - is the real challenge facing the VISTA team.

A key component of any automated control system, Mirchandani said, is the method by which the control system ``looks ahead'' to see where the road is leading.

Some so-called smart-car automation systems under testing require huge amounts of on-board computer guidance systems to accomplish this ``path planning'' task. For example, the NavLab project at Carnegie Mellon University uses sophisticated laser, sonar and stereo optical sensing systems to plot a pathway on the fly, even in off-road applications.

Another project at UC-Berkeley uses magnetic nails driven in the roadside every meter or so to chart the road path, while magnetic road strips developed by the 3M Corp. are being used to guide snowplows where drivers can't see the shoulder through the snow.

Such imbedded road markers can quickly become prohibitively expensive, said Wang, who is editor-in-chief of the International Journal of Intelligent Control & Systems.

The UA researchers are aiming for something more practical - and far less expensive - to navigate known highways and roads.

Since the paths of known highways such as Interstate 10 can be charted to within inches, the VISTA researchers are testing a ``calibration-based'' system that uses pre-loaded road-path data, along with guidance stations placed every mile or two to regularly update data.

``Even (placed) every mile, that will save a lot of time and money,'' Wang said.

The VISTA car's camera system helps keep the car in the middle of a lane. To smooth out the guidance control, Wang said the VISTA system uses ``fuzzy logic'' - a type of programming that recognizes not just ``yes'' or ``no'' values to make artificial-intelligence systems more humanlike.

While the VISTA car is far from completed, the vehicle performed well in a test in late April along an unopened stretch of State Route 51 in Phoenix, ADOT's Owen said.

Mirchandani said the system is being designed to work with pre-loaded data, perhaps from CDs, but in the future the system could use location data from Global Positioning System satellites.

Though a VISTA-like system might add thousands of dollars to the cost of a car, Wang said the control system could easily be miniaturized by creating microchips to do the job of the VISTA car's compendium of computer parts, controllers and cabling.

``In the future, we're going to put everything on a single (circuit) board,'' Wang said.

Hands off!

UA researchers making self-driving car practical, inexpensive

How it works: Smart car