Educational Media Reviews Online

John Lithgow narrates The Great Robot Race: The DARPA Grand Challenge, which asks as its central question, “can a machine navigate miles of punishing desert terrain with the speed and agility of a human driver?” Ultimately, the answer is, “Yes.” At the California Speedway, unmanned vehicles ranging from motorcycles to tricked-out semis compete to qualify for a chance to race against their ilk in the DARPA Grand Challenge. This NOVA production follows the main competitors in the race during the October 2005 qualifying event--an obstacle course--and the race itself, which, for the first time, ended with a clear winner.

The government agency DARPA, which stands for the Defense Advanced Research Projects Agency, sponsors the race, which was started in 2004, and provides the purse. DARPA, which aims to “stimulate technology that can be used by the military,” issued a challenge to construct an autonomous vehicle. At the 2004 Grand Challenge, no vehicle completed the 140 mile course. In its second year, which is shown in The Great Robot Race, five vehicles or teams finish and the challenge is met.

While many of the competing vehicles’ workings are described, two are explained in great detail and emerge as the main competitors – Stanford’s team with their robot Stanley and Carnegie-Mellon’s two robotic vehicles Highlander and Sandstorm. These well-funded teams are contrasted because of their differing approaches to the challenge. Team Stanley concentrates on software, while Teams Highlander and Sandstorm focus on the hardware. In addition, Team Stanley attempts to have the robot (a modified Volkswagen) make its own decisions based on the terrain while Teams Highlander and Sandstorm rely on intensive human data input and speed. Adding a personal angle to the robot race is the fact that the two team leaders were former colleagues at Carnegie-Mellon University.

The main problem for roboticists who are trying to construct autonomous (non-remote controlled) vehicles is translating visual images into numbers that can be made into a map and understood by the computer. This is how robots navigate from point A to point B. There are a variety of ways to achieve this and the methods and results vary. Often robots are shown in a highly controlled environment, such as a lab, moving hesitantly around large stationary objects. They have come a long way and the Grand Challenge is indicative of how sophisticated computers and robots have become.

The relevant visual images for the Grand Challenge robots are pictures of the terrain – smooth stretches, edges and obstacles - and the computer code needed to keep the vehicle on the flat spots and away from the edges and objects. Controlling the actual car functions of acceleration, steering, and braking is a given and in some cases the vehicles arrive with these modifications installed.

While it sounds simple, the problem of recognition, or seeing, is central to robotics and Team Stanley and Teams Highlander and Sandstorm take differing approaches. Teams Highlander and Sandstorm use laser range finders mounted in pairs to judge distance and terrain. Lasers fire at a fixed rate and the time it takes for the laser to return allows the robot to judge distance. Using hardware like lasers mounted on rotating gimbals to reduce vibration, Teams Highlander and Sandstorm use the strategy of gaining speed on the smooth stretches in an effort to win the race. Details of the course are magnified and studied by team members and the direction and exact course of the robot’s motion is programmed in before the race begins.

Team Stanley uses “adaptive vision” involving both lasers and video cameras which synthesize to form a color map that allows Stanley to follow the safe color path along the course. Because mapping is done on the fly, Stanley may go slower but Stanley also goes its own pace, without the human labor involved up front. Stanley is more like the romanticized notions of robots who can “think for themselves” rather than following a mechanized, rigidly defined path programmed by humans.

Spoiler: Stanley passes Highlander, a modified Hummer, at mile 122 (of 140) and wins with a time of 6 hours and 53 minutes and an average speed of 19 miles per hour. Sandstorm finishes third.

The story of the Great Robot Race is dramatized by drawing sharp distinctions between methods, styles and showing the travails of human and robotic team members – especially the low tech teams with small budgets and homemade, clumsy robots. There are times when the filmmakers have obviously set up the drama by dichotomizing the styles of team leaders and setting up timely, if poorly delivered “moments” and reactions from team members. But those flat spots are relatively few, considering that the people shown are less exciting than the robots and that they are mostly science geeks.

Putting a human face on the competition is what makes it watchable by a broader audience than those interested in robotics. But there is enough science and nice graphics included to appeal to fans who watch Junkyard Wars or Battlebots. Reminders that these robots are created for the purpose of future “unmanned war” is at odds with the humanizing story line in the Great Robot Race but the militaristic applications are generally downplayed in favor of celebrating the exciting technological achievement Stanley’s win represents.

I recommend this program for high school or undergraduate classes that cover robotics, engineering, artificial intelligence, computer-human interaction and computer design. It is very accessible with high production values and, at 56 minutes, a good length for in-class viewing. For collections that are looking for media to support summer science programs geared toward teens and technology, this is a great choice.

The entire program is available for viewing as well as a full transcript, video podcast, video extras and teacher guides and links on NOVA’s webpage about the Great Robot Race.

Additional information is available on the U.S. Government’s DARPA Grand Challenge 2005 website.