Teams Tune Up Unmanned Cars for Mojave Challenge
By MICHAEL THURESSON
The Defense Department wants to bring unmanned vehicles down to earth.
Looking to duplicate the success of unnamed aerial vehicles like Northrop Grumman Corp.’s Global Hawk, the Defense Advanced Research Projects Agency will pay $1 million to any engineer, scientist or tinkerer who develops a vehicle that can find its way unaided from Barstow to Las Vegas in less than 10 hours.
Called the Grand Challenge, the 200-mile race across the Mojave Desert commences on March 13, with teams from the California Institute of Technology, Carnegie Mellon University in Pittsburgh and Ohio State University’s electrical engineering department, along with more than 100 others. Most teams are heavy with experienced engineers and software developers working full-time jobs.
“We expect to see the autonomous ground vehicle to begin to demonstrate a level of behavior that is unprecedented today. Most robotic vehicles today are tethered,” said Air Force Col. Jose Negron, project manager for the race.
There are 105 teams listed as potential participants on DARPA’s Web site, including several in Southern California. The teams were required to submit technical papers describing their strategies to DARPA by Oct. 14. Only 20 to 30 teams are expected to be approved for the race.
Five teams, including those from Cal Tech and SciAutonics LLC in Thousand Oaks, have already been approved. The rest of the cuts will be made by early November.
The desert route, which DARPA has kept under wraps, is expected to be treacherous. The teams expect much of the race to be held within the confines of Fort Irwin, a U.S. Army training base covering 635,000 arid acres between Barstow and Las Vegas.
The lack of information has forced teams to be especially creative as they meld the necessary expertise in robotics, mechanical engineering, software development, sensor technology, satellite-based navigation and auto mechanics.
“We are looking at mobility, artificial intelligence and behavioral science in these vehicles,” said Negron, who gives teams a 20 percent chance at best to finish the race. “It’s critical that the internal mechanisms of these vehicles work together.”
The Grand Challenge will be rerun every year through 2007, with future locations yet to be decided. An award will be paid to each winner annually. If DARPA likes the results from this year’s robotic vehicle race, it will test different technologies in future races.
Universities have been the central hub of development. Cal Tech has used undergraduate students extensively while the CyberRider team in Irvine has an engineer contributing from the Massachusetts Institute of Technology.
Teams lacking strong university ties have had to rely on sponsorship or company connections. SciAutonics is comprised of engineers working full-time at Rockwell Scientific in Thousand Oaks, which supplies the team with computer equipment and software.
DARPA has left the design of the vehicles up to the teams. Some are using dune buggies, many are using converted SUVs. Part of DARPA’s review will focus on whether the cars are rugged enough to withstand the race and possess sophisticated enough navigation systems to evade obstacles.
The final step for teams after passing the November review will be a test run at the California Speedway racetrack in Fontana the day before the race. There, DARPA will run the vehicles through an obstacle course and determine if they are ready to compete.
Tales of the Tape: Several of the teams competing in the Great Challenge.
Axion LLC, Westlake Village
Makeup: Fifteen team members, including several mechanical engineering and software development students at San Diego State University. Also has digital map makers in Hawaii and at California State University Northridge, an infrared radar specialist in Westlake Village and engineering professors at Utah State University and University of Colorado.
Budget: Under $100,000
Backers: Melo Water, a bottled water company from Kosrae, one of the islands in the U.S. protectorate of Micronesia in the South Pacific. With U.S. operations based in Westlake Village, it established Axion as a separate company for the race. National Instruments Corp., a mechanical instrument design company, has given steep discounts on equipment.
Strategy: Using radar, infrared laser technology and sensors to scan the road and satellite technology and digital maps to position and guide a 1994 Jeep Grand Cherokee with 185,000 miles.
Strengths: Separation of duties plays to strengths of specialists in computerized mapmaking and infrared technology.
Weaknesses: Communication problems because team is not working together in one location.
Cal Tech, Pasadena
Makeup: Twelve undergraduate mechanical engineering students, one full-time mechanical engineer on staff
Budget: $400,000 ($285,000 spent)
Backers: Cal Tech, Ford Motor Co., IBM Corp., Northrop Grumman Corp. and various computer component manufacturers. The Jet Propulsion Laboratory gave the team image processing software. Missile navigation technology the school previously developed for Northrop is also being used.
Strategy: JPL’s software processes data received from vehicle-mounted cameras, guiding the vehicle through the desert. Missile guidance software had been repurposed to accept camera data and manage vehicle controls. Students have built a ruggedized Ford SUV sporting a roll cage with embedded sensors that collect environmental data.
Strengths: Large budget, computer industry connections and access to Cal Tech mechanical engineering and software development students with experience building off-road racing cars. University has created undergraduate classes based on the project.
Weaknesses: Dust in the desert could present a problem to vehicle cameras.
Makeup: Ten team members working an average of 10 hours per week on the project. Team leader Ivar Schoenmeyr and system designer Sinclair Fleming hold executive positions with Aquatec Water Systems and Global Telematics Inc., respectively. Other team members include researchers at MIT and UC Berkeley and a Detroit-based automotive design expert.
Backers: Personal savings of team members, equipment discounts from suppliers such as Laseroptronix (laser guidance system), Aquatec (flow control devices), CSI Wireless Inc. (satellite technology) and Group Mobile Inc. (computers).
Strategy: Uses laser guidance system that scans terrain. Data collected feeds vehicle control system that filters out useless environmental data and makes decisions. Vehicle uses satellite signals to pinpoint location.
Strengths: Blend of practical and theoretical expertise. Aquatec makes components that are used in off-road vehicles. Global Telematics makes satellite-based car navigation systems.
Weaknesses: Team has little experience working with laser sensors.
SciAutonics LLC, Thousand Oaks
Makeup: Company comprised of more than a dozen engineers from Rockwell Scientific. A California Lutheran University professor with expertise in computer vision software design is also involved. Team members donate off-hours time.
Budget: $250,000, only half funded
Backers: Bank of America Corp., Goodyear Tire & Rubber Co., Rockwell Scientific and several other computer and car equipment suppliers.
Strategy: Fielding two vehicles, a six-foot long dune buggy and a donated off-road racing car. Leveraging car navigation software designed in the mid-1990s by Rockwell that steers a car automatically as sensors detect obstacles. Testing the vehicles extensively in the Mojave Desert, at first by remote control and gradually making vehicles autonomous.
Strengths: Diversity of its vehicles and hardware and software donations from Rockwell. Reinhold Behringer, SciAutonics’ president, developed intelligent cruise control systems that prevent collisions while at Mercedes-Benz Corp.
Weaknesses: Team members can only work nights and weekends because of full-time jobs. Project needs more money to buy sensors.
Remote Intelligence Industries, Santa Barbara
Makeup: Eight full-time engineers on staff and several unpaid undergraduate students at University of California, Santa Barbara.
Backers: Using equipment and facilities at Remote Intelligence Industries, an aircraft robotics company founded by Rhandal Sayat, a UCSB engineering graduate. Team has received free consulting from other robotics companies.
Strategy: Taking the company’s expertise in making robotic aircraft and applying it to a ground vehicle. Using race as promotional tool to obtain new contracts related to robotics. If proposal is accepted by DARPA, it will continue to tap cheap labor available at UCSB, obtain vehicle sponsorship from major car manufacturer and commit more company resources.
Strengths: Close knit team with strong mechanical hardware and software engineering background and experience building government-level robotics.
Weaknesses: Lack of vehicle may increase risk of proposal being rejected or falling behind other teams. Robotic ground vehicles face more obstacles than robotic aircraft, so team might face unforeseen problems.
Compiled by Michael Thuresson