DARPA Robotics Challenge

Posted on April 20, 2012

DARPA–the Defense Advanced Research Projects Agency, an agency of the United States Department of Defense–is one of the entities which has done the most for the advancement of technology in general and robotics in particular (at least as far as publicly visible actions are concerned).
It has a long standing tradition of putting forth challenges to the community regarding some areas where it thinks some stimulus is needed to push our knowledge and capabilities further.

One good example of such challenges was the famous DARPA Grand Challenge, which consisted of a long distance competition for fully autonomous (driverless) cars, along desert roads.
The first event was held on 2004 and none of the competing robots managed to finish it.
Then, in the second event held in 2005, five vehicles (out of twenty three) completed the course.

Driverless cars are something that we’ve been hearing a lot about these days, with some of the existing prototypes having driven hundreds of thousands of miles in real-world roads and some states in the USA even handing out licenses for such robots to drive on their roads, but if we think back to 2004, when the first challenge was held, it was nothing short of miraculous for any car to drive even 100 meters without assistance, let alone the full 240km of the race.
And yet, just one year after, Stanford’s University’s Stanley did just that and did it without a hitch.

Later, in 2007, DARPA ran another competition–the Urban Challenge–where the cars where also expected to drive autonomously for a long distance, under time constraints, but this time the challenge was run on a closed “urban area-style” course, where the cars encountered other traffic (both robotic and human) and had to obey all traffic rules and regulations while navigating their assigned course.
Even if it doesn’t look like it, this was a significantly tougher challenge for the robots and in the end six team finished the course successfully (out of eleven finalists).

Just as an aside, the practice of posing grand challenges to the community is not an exclusive of DARPA. I cannot help but briefly mention, by way of example, the X PRIZE Foundation’s space initiatives which gave us the first private space flight five years ago with it’s Ansari X Prize and which is currently, together with Google, trying to send a private robot to explore the moon.

Getting back on track, the robotic community has been abuzz in the last few weeks about a new challenge coming from DARPA: the DARPA Robotics Challenge.
This time, the competition revolves around the creation and/or programming of (preferably humanoid) robots to operate in search-and-rescue types of environments.
There will be different “tracks” that teams can compete on, with some of them expected to develop both the hardware and the software and others being given the robots and asked to develop “just” the software .

To give you a taste of what the challenge entails, the robots will be expected to:

  1. Enter into an automobile vehicle and drive it to the site;
  2. Travel across a ruble-strewn path towards a building;
  3. Clear the building’s entrance of debris blocking access to it;
  4. Open the building’s door and enter through it;
  5. Once inside the building, climb up a ladder and walk across an industrial walkway;
  6. Break through a panel, with the help of a power tool;
  7. Locate a leaking pipe, identify the valve for that pipe and close it;
  8. Replace a cooling pump.

All of these tasks must be accomplished by the robots in semi-autonomous mode, which means there will be some supervision of it’s actions, but that supervision must be light, performed by non-expert operators and communication links between the robot and the operator will be spotty.

This is a rather tall order given the current state of robotics, but then that’s the whole point of these challenges: to push the boundaries of what we can achieve in a very significant way.

If you want to know more about this fascinating challenge, I would point you towards the Broad Agency Announcement (the official documentation) and IEEE Spectrum’s interview with Dr. Gill Pratt (the creator of the challenge).