 |
 |
 |
 |
| |
|||||
|
|
|
Home > News & Events > MITRE Publications > The MITRE Digest > | |||||||||||||||||||
MITRE-Sponsored Team Enters DARPA Grand Challenge February 2005
MITRE is sponsoring a team in the Defense Advanced Research Projects Agency (DARPA) 2005 Grand Challenge, a $2-million contest sponsored to accelerate the research and development of autonomous ground vehicles. The competition will test the abilities of robotic vehicles to travel over rugged terrain, guiding themselves using technologies such as radar, global positioning systems, and artificial vision. DARPA will use the most promising innovations to develop robotic vehicles that could save lives on the battlefield. Supporting the Grand Challenge competition will help MITRE develop its skills in using robotic sensors for navigation in outdoor environments. Until now, robotic research at MITRE has focused on indoor environments, where concepts were tested in robot rescue competitions. In these competitions, small robots located simulated victims and created maps of damaged buildings. For the Grand Challenge, MITRE will move up from a 25-lb research robot to a 4400-lb 2005 Ford Explorer Sport Trac, dubbed The MITRE Meteor. Although the Ford Explorer is about 176 times heavier than a research robot, MITRE's Alan Christiansen sees the challenge mainly as one of perception. "An outdoor environment is more challenging," says Christiansen. "The vehicle will have to recognize a boulder, a tree, or a road. The perception problem is significant because our vehicle will use a lot of different sensors. The difficulty is in coordinating the information from all the sensors and fusing it together so that the vehicle 'sees' a unified picture around it and makes the right response." Christiansen notes that the autonomous vehicle requires significant computational reasoning to decide how to get from here to there. "The Grand Challenge shouldn't be too difficult from that respect," says Christiansen, "because you are given a sequence of way points, or road locations, to plug into the vehicle's computer before starting. You don't have to use a terrain map and figure out a route between a start point and an end point. The primary problem is determining what obstacles lie in the vehicle's path and how to get around them. Recognizing the obstacles and reacting to them properly is the kernel of the Grand Challenge problem." Last year's course was 142 miles long, but the farthest any vehicle traveled on its own was 7.4 miles. The vehicle, a modified M998 Humvee, veered off-course on a hairpin turn in a mountainous section and lodged its undercarriage on a steep, rocky incline. The prize for the 2004 Grand Challenge was $1 million. To win, an autonomous vehicle had to guide itself over a 142-mile course through the Mojave Desert in the shortest time (and it had to be under 10 hours). Although no vehicle completed the course, there were many lessons learned, and DARPA obtained a number of important ideas that could lead to promising developments. For the Grand Challenge 2005, DARPA upped the prize to $2 million, hoping to gain more entrants and more innovations. Christiansen and his team looked at a number of possible vehicles for the Challenge, examining the tradeoffs in availability, cost, and features. The Sport Trac presented a good blend of sufficient interior room to install equipment with a relatively small overall size. And the Sport Trac's pickup-type bed may prove useful for mounting a gasoline generator to provide electrical power, if needed. The team also looked at a wide range of equipment to meet the Challenge. For example, it chose a suite of sensors that includes a Trimble global positioning system (GPS) for zeroing in on the way points. A second GPS unit will be integrated with an inertial measurement unit (IMU), which is a collection of accelerometers that measure the direction the vehicle accelerates. The combination of the GPS and IMU will provide velocity and position. "It's a system that augments GPS in terms of telling you where you are in a global reference frame," says Christiansen. "We can also do dead-reckoning if we know the vehicle's wheelbase, the steering angle, and the rotation of the wheels. Of course, this only works with no wheel slippage. "We'll also use a Honeywell HMR3000 onboard compass with roll and tilt sensors so the truck knows if it's going up or down hill and in which direction. Laser range scanners will be used to measure the distance from obstacles in the road. An onboard diagnostics module from Harrison R&D will be used to monitor what's going on inside the engine and the drive-train of the vehicle. It's similar to the diagnostic computers that auto mechanics use to check out your car. It records wheel speed, fuel level, etc." The team will also add another monitoring system from Harrison R&D that includes a Controller Area Network (CAN-bus). The CAN-bus is a serial communications bus for real-time control applications that operates at data rates of up to 1 Megabits per second and has excellent error detection capabilities. The vehicle's artificial vision will be provided by a two-camera stereo vision system. Like your own eyes, the stereo vision system gives the vehicle depth perception. The ability to measure distance visually can be combined with the radar system's output to give a more reliable distance measurement than would be possible if only one system were used. At this point, the team members are procuring the components they need for the Challenge and are integrating them into a responsive, autonomous vehicle. "This project gives team members a chance to use what we've learned in previous research projects and expand on our knowledge so that we can contribute to the development of a new type of vehicle to meet the military's needs," says Christiansen. —by David Van Cleave Related Information Articles and News Websites |
||||||||||||||||||||
| Page last updated: February 17, 2005 | Top of page |
Solutions That Make a Difference.® |
|
|
