Fontana (CA) - 23 vehicles are picked to enter the DARPA Grand Challenge 2005 race. Overall capability of the robot vehicles has dramatically improved over the previous year, but there are some teams that in fact may be capable of the winning the $2 million cash price - one of them being Stanford, which completed four perfect runs in the qualification round.
The DARPA Grand Challenge is full of robotic vehicles that entertain the crowd by making unexpected turns, crashing into walls or catching on fire. Team Stanford's (http://www.stanfordracing.org/) vehicle, Stanley, has completed four perfect runs at the National Qualification Event at the Fontana Speedway - all without hitting an obstacle or making wild turns. We chatted with Michael Montemerlo, Software Lead of the Stanford Racing Team, to find out what makes their vehicle the current leader of the pack.
According to him, Stanford chose the VW Touareg since the electronics already are tightly integrated with the vehicle and the car "is mostly" drive by wire. Wheel speeds, engine RPM, the current gear and suspension deflection are natively recorded by a "CAM bus" and Montemerlo told us it was trivial to tap into the data stream. In contrast, other teams apparently burned precious development time by hacking their own way of getting to this information.
The function of primary obstacle avoidance is accomplished with the five LIDARs (light detection and ranging) units, costing $12,000 each, mounted on the roof of the car. They can detect obstacles 30 to 35 meters away and according to Montemerlo, are accurate within 1 cm. Each unit produces 181 points of data for every line drawn with 75 lines are drawn per second. The units are focused at different distances in front of the vehicle with the closest focus distance just beyond the front bumper. "It's useless 99 percent of the time, but is necessary to see when you are cresting a hill," said Montemerlo.
The raw laser data is saved to laptop hard-drives using a compressed ASCII format that is easy to handle and edit. "The laser data stream is about 250 MByte per hour, which is not that bad," said Montemerlo. While other teams have gone with flash memory, Team Stanford decided against solid-state memory as harddrive never failed during the development of the vehicle.
Two radar units mounted to the left and right of the LIDARs. Each of the 3 X 3 square inch radars have one transmitter and two receivers, allowing the vehicle to generate range and intensity readings to a list of targets. While the LIDARs produce a terrain map of the world, the radar units help with general detection of dense obstacles such as telephone poles, metal signs and cars. Montemerlo said that they have tremendous range allowing the vehicle to see out about 200 meters (600 ft).
Sensor data then is fused with the information coming from the navigation sensors mounted on the roof of the car. A Novatel GPS receiver provides world location within 10 cm (4 inches) of accuracy. In addition, two antennas form a GPS compass that determines pitch and yaw to two degrees of error. The position estimator software combines the signals from the GPS units on top, the inertial measuring unit in the back and the CAM bus. Stanley can estimate its current orientation (xyz, roll, pitch and yaw) accurately to a quarter of a degree. Montemerlo told us that the Stanford team cares more about pitch rather than roll, "We don't worry about roll because by the time you get into a serious situation, it's too late," he said.
Training a robot vehicle to survive a 150 mile race
Team members drove the vehicle to help it learn how to avoid obstacles. As the human drove around, sensor data was collected and later used to train the software on how a real obstacle should look like. In the beginning, around 12 percent of seen objects were false positives and according to Montemerlo, the vevicle has not been driveable then. After several driving sessions, the vehicle detected 1 in 50,000 objects as false positives. While it may seem that Stanford is trying to make the vehicle more humanlike, Montemerlo told us that there are limits, "it can be taken too far, we are not trying to duplicate a human being, we are trying to build a car that drives itself."
When avoiding obstacles, the software will swerve and brake the vehicle to the "safety speed" specified by the team members. The safety speed is the top speed where the car can make any maneuver without exceeding lateral acceleration and flip over. Montemerlo told us that unexpected results happened when the team changed the safety speed from 7 mph to 15 mph. If the vehicle was already below 15 MPH, a software glitch caused Stanley to gun the engine and accelerate to 15 mph. "It was scary because it would approach a gate and speed through it, breaking any speed limits imposed by DARPA or us," said Montemerlo.
We asked the team what the most dastardly obstacle could be on the actual Grand Challenge course. Montemerlo told us that cliffs, which is are negative obstacles, are extremely tough to detect. Sensor beams go out to infinity causing the vehicle to think that there is no obstacle. In addition, high speed zones could prove dangerous because of the decreased reaction times, "If there is an obstacle in a 35 mph zone, you're going to wreck a lot of robots," says Montemerlo
Most of the software runs on Fedora Core 3 Linux, but the vision uses Debian because of a software library issue. In the back of the vehicle are six Intel Pentium-M rackmount computers connected by a Gigabit Ethernet switch. All the computer equipment was donated by Intel. "The Pentium-Ms provide a lot of bang for your buck computational wise for electrical power." With the low power requirements of the CPUs Stanford can run all the computers off of the alternator. There are redundant power supplies and the whole assembly is mounted on shocks, but Montemerlo says, "The car itself is the real shock mount."
Any component can be hard power cycled and another computer can send UDP packets that can turn anything on or off. This improves reliability and helps in isolating components for testing. "We've pushed very hard for a reliable run. If you are going to win a 150 mile race, you realistically need to run 500 miles without problems. This takes into account mean time before failure," explained Montemerlo. Before the qualification runs, Stanford took the vehicle into the desert and drove 420 miles without intervention, he said.
While Stanford doesn't want to tell anyone how fast their vehicle has gone, many people think that Stanford has a good chance of winning the Grand Challenge race. Montemerlo isn't that concerned about winning and simply wants the vehicle to survive the unpredictable and perhaps destructive desert route. "It's been a great learning experience, and I just hope Stanley will be in one piece [after the race]," says Montemerlo.
Related Articles:
DARPA selects 23 teams for the 2005 Grand Challenge
"Blue Team" runs self-righting motorcycle at DARPA Grand Challenge
DARPA Grand Challenge update #3: Interview with Team Cornell
DARPA Grand Challenge update #2: A chat with Team Mojavaton
DARPA Grand Challenge update #1: Qualification day 1 results
Workout of the Day
