Thursday, March 18, 2004
Interview: Reinhold Behringer, SciAutonics LLC
Reinhold is President of SciAutonics (www.sciautonics.com), a company which was formed to compete in the DARPA Grand Challenge. SciAutonics was among the top teams at the challange, which was held this last Saturday, and made it 6.7 miles along the course with their SciAutonics II entry, just behind Carnegie Mellon's Red Team (tops at 7.4 miles).
BK: For my readers who aren't familiar with the DARPA Grand Challenge, what is the purpose and goal of the challenge?
RB: The Grand Challenge is a competition of autonomous vehicles, driving from Barstow to Las Vegas. It is organized by DARPA with the goal to leverage from individual ingenuity. A prize of $1 Million would be awarded to the team that finishes a course of 250 miles within 10 hours. These were the initial conditions. When the Grand Challenge actually took place on March 13, 2004, the route length had been reduced to about 150 miles.
BK: How many teams are participated in the challenge?
RB: 106 teams had signed up originally; of those a subset of 25 teams was accepted to attend a qualification event (QID) before the challenge. The actual challenge event was limited to 20 participating vehicles. 15 vehicles were ultimately qualified, 13 actually participated in the challenge drive.
BK: What is SciAutonics LLC, and how are you funding your efforts?
RB: Our team members founded the company SciAutonics, LLC, for the purpose of participating in this Grand Challenge. We wanted to have a legal entity, capable of harvesting the intellectual property (IP) and providing a legal framework for operation.
Our funding came from selling membership certificates to investors. Since SciAutonics, LLC, does not have any paid employees, all the funding could be used for purchasing hardware for the vehicle.
BK: I understand you have more than two entries that have been accepted into the challenge. What kind of vehicles are you entering into the challenge?
RB: The vehicle of SciAutonics-1 is a ATV-type vehicle, manufactured by ATV Corporation. It features 4-wheel-drive with a continuous variable automatic transmission. It is an ideal vehicle for overcoming rough terrain due to its ruggedness and low center of gravity. We gave it the name RASCAL.
The second vehicle - SciAutonics-2 - originally was supposed to be a "Truggy", a desert race vehicle with excellent desert racing capabilities. We lacked, however, the funding to equip two vehicles with an automatic driving system. Therefore, we accepted the offer of Elbit Systems to substitute this vehicle by their own Tomcar vehicle AVIDOR-2004. In a partnership with Elbit, we continued to pursue two entries in the Grand Challenge, with different vehicle and sensor suite.
BK: I see that some of your competitors decided on converted trucks and humvees for their base--what drove your own choice?
RB: Our concern regarding using a truck or HMMWV was that these vehicles are relatively large. Our focus was to select a vehicle that was very agile - that is it could accelerate and decelerate rapidly - and has a small turning radius. The ATV that we chose proved to be a great choice for this challenge!
BK: What are some of the challenges you ran into meeting the March date?
RB: Our team had fallen behind the originally envisaged schedule. When the qualification began on March 8, we were not sure if we could catch up. However, with a great team effort we were able to fix some of the technical glitches, and our teams were both accepted for the challenge event. The most important challenge that would need to be overcome is obstacle avoidance at a high speed. Achieving high speed is not the issue - our vehicle can follow a set of waypoints very robustly. But to detect obstacles reliably and to react correctly in time for avoiding a collision is a challenge that generally has not yet been solved.
BK: Did you think it was feasible a team might win the $1M award this year?
RB: Ok, the race is over, and it is easy to say now that no team had a chance... but in earlier statements I made the assessment that the chances of a team actually finishing this course is 1:1000. The technology is just not yet mature enough to function reliably under these racing conditions, and although there are very sophisticated and valid technical approaches, they have to prove themselves in situations other than simulation.
BK: Finally, what has been the biggest technical challenge to making this project work?
RB: In our RASCAL the biggest challenge was to work against the clock. Time was critical, and for a team like us who was working not during the day-time job but during nights and weekends, this proved to be a big issue. The largest technical challenge was to develop a suitable method for avoiding collisions with obstacles. This module took the longest to develop, because it was composed of several sub-modules: sensor communication, data fusion of several sensors, obstacle detection, path planning.