The following post is a summary of posts on Team Robomonster, a 2005 DARPA Grand Challenge entry. The project “team robomonster” managed to make a “rock crawler” vehicle drive-by-wire, and respond to the computer. It failed to get GPS input to accurately steer the vehicle, and was eliminated at the second round of the “Grand Challenge”. The 2005 Grand Challenge winner came from Stanford, and, like “Robomonster” did not try to use vision to steer.
PART I – TO THE DARPA VISIT
Wednesday, May 18, 2005
Our Site Visit…
Well, it’s been one week since the site visit for team Robo Monster, so I thought I would post a summary. Overall, there was a lot of great stuff, particularly for a team which started a mere 12 weeks ago. The DARPA reps were particularly interested in the custom-built vehicle, and talked to offroad racing engineer Brian Kirby for about an hour. They were impressed with his techniques for keeping his custom-built vehicle cool in hot environments, and his unique modular method of integrating robotic components with the base system.
The DARPA reps were also impressed with our “sensor-dense” system incorporating multiple IR and ultrasound detectors, developed by our microprocessor experts Michael Wilson and Kerstin Gilg. Working under deadlines, Michael and Kerstin created a custom board managing and processing information from 4 Sharp IR sensors plus one Devantech SRF04 ultrasonic sensor. Currently, these sensors are on the front bumper. We plan to expand the system to handle 8-15 sensors, and array the sensors in a grid on individual body panels of the vehicle, providing a sort of infra-red “touch” sense. We’re in good company in this “sensor-dense” concept – NASA recently promoted a new system they are using to make robotic arms sensitive to touch in their environment using similar IR sensors. Check out the NASA project at this link.
Despite the successes, we had some areas which are more challenging. The biggest one was our motor controller, whose Flash memory became corrupted during an upgrade and could not be used to control our steering motor. As a result our site visit was ‘static’ – we demonstrated vehicle systems in place, and did not attempt to send the vehicle to our RDDF waypoints. The Flash memory became corrupted via a series of unlikely accidents. Two days before the visit, we discovered that our motor controller (Roboteq) would need the upgrade to properly position the steering. When the basic upgrade methods did not work, we called the developer in Switzerland and walked through a special upgrade – an upgrade which, unfortunately corrupted the Flash memory instead.
Fortunately, we’ve received a special program from the developer which will rescue or motor controller – so we plan on doing a demonstration of waypoint navigation in the next couple of weeks after re-installing the software.
Other projects – Taos, Inc. is sending us a special ambient light sensor for evaluation on our vehicle. We plan to use this special sensor, which has a response range comparable to the human eye (40,000:1) in developing our “sun sensor.” This custom sensor will use the Taos chip to determine absolute lighting levels around the vehicles, and also, via a unique “sundial” arrangement, the relative contrast of shadows. Combined with another Taos sensor which measures overall color (we will use it to measure the “blueness” of the sky) we will have shadow contrast and direction information. This information in turn can be used by our vision processing algorithms to predict how deep the shadows should be around an object, as well as their location – improving Robo Monster’s vision
Tuesday, May 10, 2005
Making it happen
Good news today – our E-stop is fully functional! We’ve got compressed air that slams the brakes when the power is cut, and manual and electronic relays are installed. Computer integration has been a bit more of a challenge – yet another computer blew up on us onWednesday(!) But we’re testing our home brew laser rangefinder on a variety of colors and textures – it will be lots of fun if this $100 instrument does the job. We’ve also hooked a “fail-safe” ultrasound unit from Sports Imports, Ltd. This system was designed for parking cars, but two of our programmers have now “hotwired” the system so we can connect it directly to our computers. Our robotic senses at the site visit will be primitive but functional. – Pete Markiewicz, Team Leader
Tuesday, March 08, 2005
Past the ‘Drive by Wire’ Hump…
Well, it’s three days before the DARPA March 11th deadline for part 3 (technical description of vehicle plus video of vehicle) and we’ve done it. Last week we got drive-by wire for steering and throttle on Robo Monster, with brakes ready. We shot a video showing the ‘Monster going over 4-foot obstacles while being steered by remote control – fun! We also sent in our part 3 document describing what we intend to do with Robo Monster.
The only thing missing was the transmission – the current vehicle has a manual transmission, and an automatic transmission will be purchased during the next week or so. Until then, we can’t install a linear actuator to put the vehicle into full drive-by-wire.
We’ve decided to use standard R/C control for remote-control driving as we develop our autonomous system. The reason is that there is a fairly simple circuit for switching between our Roboteq motor controllers and an R/C control. We could do wi-fi, but looking at the structure of Roboteq-computer communication convinced us that it would be a detour.
Currently testing sensors using microcontroller boards from BasicX and Micro Basic. We plant to put lots of “point sensors” all over the vehicle and read them with standard hobby microcontrollers and software. Our vehicle will be more “tactile” than others, though we will put in a high-res visual system later.
For now, our current efforts are to build a GPS waypoint following into Robo Monster. We have a good, if tiny GPS/DGPS that we’re using for testing. Currently deciding if we want an intermediate controller (e.g. acroname brainstem) between the PCs and the motor/servo controllers.
Monday, February 14, 2005
This is the week that the vehicle goes from being a regular car to full drive-by wire. We’re installing Roboteq hardware for steering and throttle, along with some motors.
More members, this time coming from Cal Poly. We’re working out how to make the large sensor grid resistant to damage.
More sponsors this weekend…
Thursday, February 03, 2005
Team Robo Monster™
Welcome to the Team Robo Monster Wblogger! We are working on an entry for the 2005 DARPA Grand Challenge that will uniquely combine navigational ability with “attitude” in a kick-butt, unstoppable, off-road vehicle package. We have gotten our vehicle and are currently seeking sponsors and team members.
Team Robo Monster™ Responds to DARPA Grand Challenge
Desert Field Test of Robotic Vehicles Offers $2 Million Prize
Los Angeles, California, January 31, 2005 … Team Robo Monster™ announced it will participate in a Defense Department research and development initiative aimed at advancing robotics technologies for future military use. The initiative, known as the DARPA Grand Challenge, is a field test of fully autonomous ground vehicles to be conducted in the Mojave Desert on October 8, 2005. The Defense Advanced Research Projects Agency (DARPA), is offering a $2 million prize to the vehicle that completes the course the fastest within a 10-hour period.