Welcome to GISWeekly! This week's story is about process, not simply the end result. The DARPA Grand Challenge (Defense Advanced Research Projects Agency) set out to award $1 million in cash to the robotic vehicle that could complete a 150-mile course across the Mojave desert on Saturday, March 13. The race consisted of 15 entries of driverless robots that all either broke down or withdrew.
GISWeekly examines select top news each week, picks out worthwhile reading from around the web, and special interest items you might not find elsewhere. This issue will feature Industry News, Alliances/Acquisitions, Announcements, Appointments/Resignations, New Products, Going on Around the Web, and Upcoming Events.
GISWeekly welcomes letters and feedback from readers, so let us know what you think. Send your comments to me at Managing Editor
Susan Smith, Managing Editor
Bots Take Up the DARPA Challenge Using Geotechnology
By Susan Smith
This week's story is about process, not simply the end result. The DARPA Grand Challenge
A description from a press release describes the race: “Two of the entries covered about seven miles (11 kilometers) of the roughly 150-mile (240-kilometer) course in the Mojave Desert while eight failed to make it to the one-mile (1500 meter) mark. Others crashed seconds after starting. The Defense Advanced Research Projects Agency was sponsoring the Grand Challenge to foster development of autonomous vehicles that could be used in combat. Defense officials foresee using the driverless, remote control-free robots to ferry supplies in war zones.”
A great race is not necessarily all about winning. In the case of the DARPA Grand Challenge, it was the first of its kind. Technology was used to develop these vehicles and to guide them - navigation systems, GPS, satellite imagery for the maps, microcircuits and sensors - over a 150-mile course in less than 10 hours.
One competitor said the goal wasn't necessarily to complete the course.
"It's a tough challenge -- it's a grand challenge -- you can always bet that it's not doable. But if you don't push the limits, you can't learn," Ensco Inc. engineer Venkatesh Vasudevan, said.
“Today was a most important first step in a long journey,” said Dr. Anthony Tether, Director of DARPA. “Although none of the vehicles completed the course, and we were not able to award the cash prize; we learned a tremendous amount today about autonomous ground vehicle technology. Some vehicles made it seven miles, some made only one mile, but they all made it to the Challenge, and that in itself is a remarkable accomplishment.”
Even the head of Carnegie Mellon's Red Team said there was only a 40% chance they would complete the course. It was a very hard course in the beginning and lightened up as it went on. The Red Team vehicle, Sandstorm, crashed on a switchback after traveling 7.4 miles into the desert. It high-centered on the berm of a road, and the collision broke the vehicle's front pair of half shafts and shredded both front tires, then sprung a fuel leak.
According to Red Team leader, Carnegie Mellon robotics professor William L. “Red” Whittaker, Sandstorm's navigation system had never fully recovered from a spill the vehicle took the previous week during a practice run at the Nevada Automotive Test Center (NATC). Whittaker was especially pleased with the performance of the mapping system.
DARPA, the R&D arm of the Pentagon - who has been responsible in part for the invention of the internet, the UAV, and other amazing advances-- has been asked to make a certain percentage of the vehicles used during war autonomous and robotic by 2015. (The UAV incidentally is not autonomous--it's flown with a joystick.) The task for these 'bots' was to get from point A to point B, by being smart about what terrain they were covering and what speed to travel. “There is no driver,” said Gary Napier of Space Imaging, a company which sponsored the Red Team by providing 10,000 square kilometers of imagery for mapping. “You literally hit a 'go' button and the thing crosses the finish line 200 miles later or it crashes. There is no in between. It has to follow a specific course that DARPA determines and provides to these race teams only two hours in advance.” The teams know generally where the course is, but in that two hours they're going to be given 1,000 GPS points. The robotic vehicle must hit all the GPS points going through some sort of corridor, and traverse many different terrain features, under tunnels, etc.
This is where GIS technology comes in. Maps of the environment were designed specifically for the vehicles. Sandstorm had LIDAR projected in front of it, RADAR and stereo cameras to detect objects that it had to avoid, all done in real time. The map was used to tell the vehicle where to go to the next GPS point and to understand how fast it can go according to the type of terrain that was predetermined in the map, using real time information to understand if a boulder was in the way, or a gully was washed out since the last time the maps were made or if there was truly a barrier there, or another race vehicle to go around.
The imagery for the Red Team was integrated into proprietary SAIC software. They also used software from a company that spun out of Carnegie Mellon called TerraSim.
All six of the teams using NavCom's StarFire Network for the navigation of their autonomous ground vehicle in the race qualified to compete in the $1 Million DARPA sponsored Grand Challenge event. NavCom's StarFire Network and receivers are part of a global satellite based augmentation system (SBAS) that provides precise positioning accuracies within a few centimeters of truth.
DARPA had informed contestants of the general large area between Barstow, California and Las Vegas, but then changed the destination to Primm, Nevada, right on the California/Nevada border. “They wouldn't tell anything else until two hours before the race, so they used satellite imagery and USGS DOQ aerial imagery--that's quite old but has some pretty good accuracies to it--to develop these terrain maps and develop predetermined corridors, the best possible corridors that could be used.”
“There were 20 people devoting thousands of hours on Intel computers to come up with the best possible corridors, then they would run simulations against them to determine speeds. Intel, Boeing and SAIC are all large sponsors of this - Seagate and ATI graphic cards, Intel donated all the computers for the mapping crew,” said Napier. This mapping crew took a 23-foot construction trailer out to the starting line. Twenty people were in there when they got their mapping waypoints two hours ahead of time, and in a mad rush developed and extracted the route and adjusted it into Sandstorm. In that two hours they fine tuned maps that they had developed over the past three months.