|
Getting your Trinity Audio player ready...
|

The research done by people like William Whittaker, Ernst Dickmanns, and many other researchers around the world was essential for DARPA’s next step. The ground was prepared to accelerate the research progress. It was time to find the best people to put driverless car technology into the hands of the private industry.
The New Era of Autonomous Vehicles began in the Mojave Desert
So far, these ‘driverless cars’ still require a driver – at least for handling emergencies. Most would fit into the category of Advanced Driver Assistance Systems (ADAS), but not fully autonomous just yet. They were doing great with the simple task of driving along a highway and staying between marked lanes, but there was still plenty of human attention and interaction required in many urban settings. A lot of research parts were already available, but DARPA wanted to accelerate progress and create actual autonomous vehicles capable of adapting to all scenarios of peace and war.
At DARPATech 2002, they announced the DARPA Grand Challenge 2004 – a race for fully autonomous vehicles. They challenged anybody interested in coming up with self-driving vehicles that would drive themselves from L.A. in California to L.V. in Nevada.
… and with L.A., they meant Barstow, and with L.V., they meant Primm, and instead of streets and roads, they would be driving across the Mojave Desert.
2004: GhostRider vs Sandstorm
“To finish first, first you must finish” [Zarathustra]
The favorites to win this challenge were Carnegie Mellon University (CMU) and its Red Team, with William “Red” Whittaker and Canadian Chris Urmson driving the technology. They got $500,000 in cash donations and millions more in donated tech and man hours from various sources (Caterpillar, Goodyear, Intel, Boeing, and a small search engine company named Google, ….). This was all brought together in a red Humvee called Sandstorm. Sandstrom had GPS, stereo vision, radar based on radio waves, and lidar based on infrared light.
French-American Anthony Levandowski and his Blue Team from UC Berkeley chose a motorcycle, which cost an estimated $100,000 to build and was named GhostRider.
The race was supposed to follow a 142-mile course and take 10 hours, and whoever achieved it in the shortest time would take home the price of one million dollars. Since this is America, one competitor wanted to install a roof-mounted cannon to win. Fortunately, the cannon was not necessary as the race rules weren’t ‘robot vs robot,’ but ‘robot vs desert’ … and the desert won each and every time. None of the vehicles actually made it to Primm. CMU’s Red Team – thanks to its NavLab advantage – was still the best team reaching 7.4 miles.
GhostRider didn’t even make the start. Levandowski forgot to turn on the gyro, which is supposed to keep the motorcycle upright. The robotic two-wheeler was still part of history and eventually made its way into the Smithsonian. Some would call the race a debacle. But others compared it to America’s space race and called it a ‘moonshot’ event.
2005: Stanley versus H1ghlander
Nobody won the $1,000,000 offered for the 2004 Challenge, so DARPA doubled down. DARPA set the prize money for the DARPA Grand Challenge 2005 at $2,000,000 for the winning team. Almost 200 teams wanted to participate, but only 23 finalists qualified. In the end, only five cars reached the finish line of the 120-mile course.
This time industry giants got even more involved as sponsors. Thanks to GM, CMU got to start their old Humvee Sandstorm and a newer one named H1ghlander. The Silicon Valley VW Lab donated a VW Tuareg to the Stanford Team and the tools to hack into the car’s electronics. While Challenge 2004 was all about finding the proper hardware, sensors and vision, Challenge 2005 was more about improving the software.
CMU was still the favorite, but two former CMU employees had moved to Stanford, which entered the competition for the first time. One was Mike Montemerlo, and the other one was German AI scientist Sebastian Thrun. At CMU, he helped develop the robots that would map old coal mines in Pennsylvania. Thrun created a small team of Stanford students and engineers to create Stanley – The Volkswagen. Stanley beat out 23 other vehicles and finished in under 7 hours. The two CMU Humvees, Sandstorm and H1ghlander, came in second and third. But really, there were 23 finalists that day, which achieved what many thought impossible.
2007: Junior vs Boss
But let’s not forget that DARPA is a military agency and, as such, hasn’t been satisfied yet. None of these vehicles would be able to navigate the dangerous urban environments of Iraq or Afghanistan. So, in 2007, they promoted the DARPA Urban Challenge. The task was to obey all traffic rules of the California Drivers Handbook, interact with other traffic, no human driver and no interventions allowed. The vehicles also had to learn the most complicated traffic elements known to humans: The left-turn into heavy traffic, the American All-Way Stop, and dealing with human drivers breaking the rules.
Some 138 teams showed interest, only eleven made it through all rounds of eliminations, and only six completed the 60-mile course.
Stanford Racing wanted to defend its title against CMU’s Tartan Racing but didn’t put too much effort or funding into it. CMU was again the favorite due to its funding and deep connections to GM. This time CMU’s 2007 Chevy Tahoe Boss took the top price, and Stanford’s Junior was second. The other successful teams came from Virginia Tech (“Odin”), MIT (“Talos”), the University of Pennsylvania (“Little Ben”), and Cornell (another Chevy Tahoe named “Skynet” – this name should come with some dark, foreboding music).
Post.Scriptum
These researchers and scientists, with their Stanley, H1ghlander or Boss, were still lightyears away from what Dr. Gustav Stuempfel or Knight Industry had achieved decades earlier. When inspecting vehicles during the elimination round, DARPA’s program manager apparently said, “Boss behaved like a good beginning driver.” Others compared these cars to being inebriated.ย These cars were apparently driving around like drunk teenagers with real speed regulators.
More Information
- Team Blue GhostRider
- CMU Autonomous Vehicles
- 2004 DARPA’s Debacle in the Desert
- 2004 DARPA Grand Challenge
- 2005 DARPA Grand Challenge
- DARPA Urban Challenge: Final Event Highlights
Editorโs Note: The views and opinions expressed in all blog posts are those of the authors and do not necessarily reflect those of the Redwood City Pulse or its staff.



