The DARPA Robotic Challenge Finals, a 3-day affair which just concluded at the Pomona Fairgrounds, is a far bigger event than I imagined. 24 teams participated and 3 won the $3,500,000 prize money. Team KAIST from South Korea won.

Dr. Gil Pratt headed the 2-1/2 year challenge and staged the event at the fairgrounds. The overall objective was to stimulate awareness and interest in the development and use of robots to help with disaster relief. The Fukishima nuclear disaster was the point of origin: no robots were immediately available to provide assistance nor were any available for months after. Hence the challenge: provide robots that could assist in future disasters — robots that could turn valves, move over and around rubble, handle and use power tools, and operate semi-autonomously particularly in areas with poor or no communications.

DARPA built four identical testing sites at the fairgrounds to the one shown here so that each robot team could be tested in semi-real situations to perform thos tasks.

If you study the picture you can see the whole process that each of the 24 participating teams went through:

1. The robot got loaded into the drivers seat of an AGV under the white awning. It then drove itself down an obstacle course to the red door at the bottom right of the picture.
2. Then the robot dismounts the vehicle.
3. I then opens and goes through the door.
4. Then it moves past some electrical panels and finds a valve which it must turn sufficient to close the flow.
5. It then moves over to the black circle and chooses a tool or tools from the shelf to cut around the circle in the drywall.
6. It then returns the tools to the shelf and walks or climbs over the rubble through the doorway to the stairs.
7. A special task was thrown in; on the day I watched the task was to locate a switch box and pull the red lever down.
8. And the final task was to climb the stairs.

For each task the robot was able to do it got a point; 8 points being the best. The three teams with the highest point counts got the big bucks: $2 million; $1 million and $500,000. Because all three winners had the same point count, the fastest teams won. Here’s the point count and times for the 3 winning teams:

1.  Team KAIST  44:28 minutes, 8 points
2.  Team IHMC Robotics  50:26 minutes, 8 points
3.  Tartan Rescue  55:15 minutes, 8 points

There were 4 teams that got 7 points. Each of the 24 teams had at least 10 members; some upwards of 20. All were well-educated roboticists supported by research labs from major universities and, in a few cases, corporations. And all were busy making their solutions to the set of tasks work — doing the software, control work, hardware development and communications.

Some of the robot solutions were quite inventive. For example, the Korean winning team robot could both walk and roll. It could transform itself from a standing position used for walking to a kneeling pose meant for wheeled and fast motion. Another example was the 3rd place team from CMU (called CHIMP). It rolled on ruberized tracks like a tank, driving on all fours over obstacles then stands up on its hind legs to provide a stable platform for precise manipulation.

Nevertheless, each test run was supposed to happen in less than 60 minutes which, because of penalties and falling down robots, took much longer, the action was slow. As technology writer John Markoff wrote in the NY Times:

Reporters were once again left grasping for appropriate metaphors to describe the slow-motion calisthenics performed by the menagerie of battery-powered machines. Most agreed that “like watching grass grow” was no longer the best description, and Gill Pratt, the Darpa official in charge of the competition, suggested that it had risen to the level of “watching a golf match.”

This is a big story with implications well beyond the contest aspects. It reflects the state of the art today in humanoid robotics and shows that we have a long, long way to go to have truly helpful disaster relief robots.

DARPA Robotics Challenge official Brad Tousley said, “There is a long way to go. There’s fact and there’s fiction. There’s a lot of fiction out there that robots are much more capable than they really are. But part of DARPA’s job is to show the possible, and what we can start to do. And then often other organizations, and other countries or other companies will invest more to bring it along. But it’s our job to start that process.”

The overall experience made me appreciate the DARPA challenge and the money invested to get the ball rolling in this important area of disaster relief. But I also wonder over the complexity and capabilities of the human body and brain which the challenge attempted to emulate. Ray Kurzweil suggests that 2045 will be the point where technology will surpass human and particularly brain capabilities. Perhaps… perhaps not.