Click Here
for more articles |
|
|
Great competition failure or victory? |
by:
Yuri Iserlis |
FACTS: As everybody knows, the DARPA Grand Challenge ended with the failure on March, 13 because not one robot vehicle could pass the 142 mile course in the Mojave Desert. One hundred and six teams originally applied; that number was shortened before the starting date; and 15 teams made it to the starting line in tiny Barstow, hoping to reach the finish line in Primm, Nevada. The exact itinerary was a secret until about three hours before the start of the race. In reality, the course included unpaved utility roads, switchbacks, uphill and downhill slopes, sharp turns and cliffs.
GOAL OF THE RACE: The idea to make unmanned vehicles competition based on the Congress decision that one-third of military vehicles must operate unmanned by 2015. Defense Advanced Research Projects Agency (DARPA) as the central research and development organization for the U.S. Department of Defense, held the Grand Challenge Race to accelerate technological development for military applications, to receive new ideas and check some ways of the development of the given task. DARPA decided to dangle the big prize of 1 million dollars in front of small and big companies, and even the “Average Joe” to see if they could come up with cutting edge ideas in the field.
BACKGROUND OF THE RACE: This competition was organized under DARPA supervision not from scratch. Achievements of AI science and technical level of modern industry prepared ground for this Race. It makes sense to recall some innovations:
• in the field of navigation systems as GPS system, Distronic autonomous intelligent cruise control (Mercedes-Benz) ,
• in the field of all-terrain vehicles as the military ATV’s (All Terrain Vehicles) from Jeeps to tanks, Russian Lunokhod 1 (1970) or two of NASA’s space robots Spirit and Opportunity,
• in the field of control systems as the control systems of the existing moving robots from cars to dancing humanoid robots.
RULES: It had originally been proposed that any team could participate in the March 2004 race, if its technical paper passed muster. The vehicles could come in all shapes and sizes from ATVs to Hummers without any drivers. Each robot must be led by a series of GPS coordinates and equipped with an emergency stop device. The vehicles only had to work twice, once at the Qualification Inspection Demonstration and then on the race route. There were very few specifications for the vehicles so teams had written their own by guessing what problems they will come across. DARPA changed rules several times and cut several times the number of participants.
RESULTS: Fifteen autonomous ground vehicles were allowed to navigate a very complex route of the Mojave Desert. The first DARPA Grand Challenge concluded without a winner. Two qualifying teams declined to even compete. Only seven vehicles managed to travel at least one mile. Some teams ran into obstacles. Others had equipment such as brakes or navigation systems break down. More than half of vehicles had failures in their navigation systems. Although Sandstorm of the Red Team completed the most distance at 7.5 miles, the vehicle went off course and got stuck on an obstacle. The vehicle was disabled by race officials.
REASONS OF THE RACE’s FAILURE: For the failure of such a great competition there must be found great reasons. In my opinion, here are the following 3 main reasons, based on methodology of design, choosing of team members, and the management of the competition.
1. Methodology of design. Revealed mistakes could be eliminated, if the participants used completely for design: a system approach, thinking aid tool, tools for digital testing, design optimization tools, and a knowledge base of military standards. Thinking aid tools (see for example, www.ideationtriz.com) together with CAD/CAM/CAE systems have to not only assist designers and engineers inventing machines by generating and synthesizing new ideas, but also to help resolving difficult problems by eliminating technological barriers and other contradictions, including analysis and elimination of existing failures with prediction and prevention of potential failures. Digital simulation tool is a complex of different types of simulations and computer based research applications. For example, the ESI Group (www.esi-group.com) has developed a broad range of virtual test applications. Typical examples are aero-dynamics, aero-acoustics, electro-magnetics and bio-mechanics tests, and also crash test, drop test, stamping, casting, welding, and so on. The design optimization tool is a complex of programs for optimization of solutions on the basis of modern mathematical methods. These programs are developed for example, by Frontline Systems (www.solver.com) that has a technology platform for solving optimization problems. Some teams didn’t know about military shock and vibration standards, and they at the last minute swapped out their conventional storage devices for solid-state e-drives.
2. Choosing of team members. There are very important roles of management and human resources. I am sure, that if Albert Einstein applied to get a position in many present American companies, in 90 % cases he would not have been hired when he was young, and in 99.9% cases – after reaching 60 years age, because many companies find age, background, knowledge of famous tools, more important than brain ability. Some middle and big American companies are using primitive robots for selection of future employees with such criteria as keywords. Many software companies got rid of algorithm engineers in behalf to programmers. But a programmer by definition is not an engineer, he (she) lives in the world of codes and doesn’t have an engineer’s horizon. Any brilliant idea cost much more than expenses on teaching to programming skills.
3. Management of race. We can see some insufficient criteria for access to race, and some wrong rules of race.
• It is obvious, that military unmanned vehicle of the 21st century must have 2 main characteristics of reliability: survivability and ability to self-restoration. It means that if one part of the vehicle or device for different reasons will be out of order, other parts must compensate for it. For example, if a vehicle has 8 independently powered wheels, it can execute its main function –moving, even if it will have 3-6 wheals, depending on the position of the centre of gravity. The same must apply to the navigation system, control system and so on. The principle of redundancy has been used in military devices a long time. What self-restoration means can be clear from such a simple case when a vehicle by any reason (for example, near explosion) is turned over. In the war conditions a robot vehicle must return itself to a normal position by such ways as, for example, most of bugs in nature do it.
• If one or all the vehicles pasts the route in the Mojave Desert, it would not be a complete test for military robot vehicles, because such type of devices must function not only in stone desert and hills, but also in sand deserts, tundra , forests, jungles, swamplands , and in any climate.
Some vehicles were disabled and got out of the race for occasional reasons: broken axes, wrong position of tumbler and so on. It is obvious, that the main goal of such a competition as Grand Challenge must be revealing perspective solutions that can be used for the creation of the future USA military unmanned robots. Very strong rules of competition didn’t allow revealing completely these solutions. Therefore it would be better to give any participant a permission to use as minimum, 2 vehicles to exclude any occasional, collateral reasons and have even the right to repair any devices in the vehicles including the navigation system, certainly to have been out of route. Later this permission could be evaluated by some criteria to find a winner. This practice could be worth $1 million dollars, because a mistake at the beginning of any project would be cheaper in 100 and more times, than in the end, in the stage of the shipping to the Army.
WHAT DARPA FOUND AND WHAT IT LOST: The greatest achievement of DARPA was permission to any citizen and not citizen of USA to participate in the Defense of the country not only as soldier or taxpayer, but as student, engineer, scientist and manager. The goal of DARPA was “to leverage American ingenuity to accelerate the development of autonomous vehicle technologies that can be applied to military requirements.” DARPA solved this problem. It was not easy, because this Agency had to pay attention to many parts of USA populations, from hippies to lovers of nature. Success of the Red Team and other participants demonstrated the principal possibility of realizing a main goal – creation of some prototypes of military unmanned robot vehicle. What the DARPA lost? Some participants could not show their vehicles that could be winners. On the other hand the teams are spending their own time and investors’ money, and some of them would not participate in future competitions, because they (or their investors) lost funds or trust to DARPA.
FUTURE CONTESTS: The Defense Advanced Research Projects Agency has scheduled the second autonomous vehicle race for October 2024. For this round DARPA, is offering a $2 million prize, doubling the $1 million unclaimed prize of this year’s race.
In the meantime, a second 2004 Open Challenge has been set up by International Robot Racing Federation (IRRF) for September, 27- October 02, 2004. Goals of this contest differ a little bit from the goal of the DARPA Race. (see: www.irrf.org). This contest is also planned to take place near Las Vegas, and the organizers have set up 2 prizes: Grand Prize ($1,000,000) and Precision Prize ($250,000).
Main task of this article is to help future teams successfully win the new contests.
For more information about Artificial Intelligence and Robots see Web site: www.cleverace.com
Dr. Yuri Iserlis
Clever Ace- President
www.cleverace.com
About the author:
Some information about the author. Graduated St-Petersburg Polytechnic State University as mechanical engineer, and post graduated St-Petersburg Marine Technical University. Devoted 10 yeas to a methodology of optimization of design and post-design development for complex machines. Worked as designer, researcher, analyst and scientist in different fields of industry, education and medicine. Created from scratch the Intellectual System, Inc. in St-Petersburg (Russia - 1992-1996). Developed some real-time intellectual systems for industry. Wrote and published more than 60 articles and books. In 1996 immigrated to the USA. In 2002 founded a new company Clever Ace (San Jose, CA –www.cleverace.com)
Circulated by Josepi Group
|
|
