This past weekend, eight robots distinguished themselves from the rest at the DARPA Robotics Challenge, earning them a place at the 2014 Grand Challenge. Here's what you need to know about these great eight.
DARPA launched the challenge as a way to encourage the development of robots that can work with people as they respond to both natural and human-made disasters. The competition consists of both robot systems (Class A) and software teams (Class B and C) — and it's designed to be challenging.
When the competition first started in June, there were 100 robots. The recently concluded competition, held at a racetrack complex near Miami, featured 16 teams, of which eight qualified for the finals. The winning robot, which will be decided at the end of 2014, will earn a $2 million award.
Here are the top eight robots:
The SCHAFT robot (or S-One) is made by SCHAFT of Japan — now one of Google's primary robotics research labs. During the competition, this bot won four out of eight tasks, including terrain, ladder, debris, and hose. It stands at a compact 4'11" (1.48 meters), weighs 209 pounds (95 kilograms), is gangly limbed, 2-legged, and is notable for its use of a capacitor to drive high-voltage liquid cooled motors rather than a battery and normal servos, allowing for stronger and faster muscles. As the overall winner, S-One was particularly strong when it came to tasks requiring dexterity and mobility.
Team IHMC Robotics, Robot Atlas‐Ian, lead organization: Florida Institute for Human & Machine Cognition (20 points)
Finishing in second place was the the Florida Institute for Human and Machine Cognition's Atlas, or Atlas-Ian. The IHMC is a Track B team that's focusing its efforts on software development, namely walking and manipulation code. Built and supplied by Boston Dynamics, the six-foot robot excelled at navigating a 25-yard course consisting of three piles of cinder blocks. It took the 330-pound robot about 30 minutes to do it — but it was the best of the competition.
Team Tartan Rescue, Robot CHIMP, lead organization: Carnegie Mellon University, National Robotics Engineering Center (18 points)
CMU Highly Intelligent Mobile Platform, or CHIMP, is a human-sized robot designed to perform tasks like climbing ladders, driving vehicles, and closing valves. The Carnegie Mellon designers describe it like this:
With four limbs and a head-like sensor array, CHIMP will look vaguely simian. But the way it will move around — on rubberized tracks at the end of each limb — is more comparable to a tank.
CHIMP normally will travel with all four of its tracks on the ground, although it can stand upright and move on the treads of just two limbs when it needs to turn valves, operate power tools or perform similar tasks...The robot is five-feet, four-inches tall (four feet when crawling) and 2 1/2 feet wide, but will weigh just more than 400 pounds.
Software developers have been writing code to control the robot for months, using a stand-in robotic arm and computer simulation for testing. The robot will be able to perform the demanding tasks of the DRC through supervised autonomy - receiving high-level commands from a remote operator while employing its own intelligence to maintain stability and keep itself from harm.
The remote operator will be able to visualize and map the environment where the robot is working using immersive 3D sensing and perception systems developed by the NREC. The robot will be able to perform certain tasks autonomously.
Team MIT, Robot Atlas-Helios, lead organization: Massachusetts Institute of Technology, Computer Science and Artificial Intelligence Laboratory (16 points)
This is another iteration of Boston Dynamics' Atlas, this one driven by algorithms and user interfaces designed by a team from MIT.
Even more than CHIMP, NASA's RoboSimian is designed to imitate a non-human biped form. NASA describes the "simian-inspired" robot like this:
This groundbreaking RoboSimian uses deliberate and stable operations to complete challenging tasks under supervised tele-operation while in a degraded human environment. This emphasis on stability over dynamics and deliberation over reaction will result in faster and more robust overall operations by decreasing missteps for both robot and human operator. To accomplish this end, the team will employ design methods, system elements, and software algorithms that have already demonstrated success for JPL's unique ATHLETE and Lemur robots, JPL's first-place DARPA ARM-S team, and Stanford's selectively compliant DARPA ARM-H design. In particular, RoboSimian will use its four general purpose limbs and hands, capable of both mobility and manipulation, to achieve passively stable stances; create multi-point anchored connections to supports such as ladders, railings, and stair treads; and brace itself during forceful manipulation operations. Risk is further decreased through heading-agnostic perception, mobility, and manipulation in which costly reorientation steps and body motion are eliminated through the axi-symmetric distribution of the limb workspace and visual perception.
Another B-Track team operating an Atlas robot, this one by TRACLabs.
Team: WPI Robotics Engineering C Squad (WRECS), Robot Atlas-Warner, lead organization: Worcester Polytechnic Institute (11 points)
Called WARNER, WPI's robot was the only the only version of ATLAS to successfully complete the driving challenge. The team members of this C-Class team have expertise in robotics software, balancing and walking behaviors, robot perception, and professional software engineering.
Lockheed Martin Advanced Technology Laboratories created a version of Atlas that rounded out the top eight. The company is focusing on autonomy and control software that can guide robots.
And for those who are curious, NASA's Valkyrie finished at the bottom with zero points.
Images: Department of Defense/DARPA, NASA,