The DARPA Robotics Challenge (DRC) is finally here, and we’re here to fill you in on all of the details. Such as, what is the DRC? Who will be competing? And most importantly, what do they plan on bringing to the table? Twenty-five teams will be competing for the grand prize of $2 million starting today, June 5th. You can view the livestream here.
The competition consists of 8 tasks which involve humanoid robots performing tasks which would be necessary during the recovery efforts of a disaster. An event that motivated the DRC was the unfortunate Fukushima Daiichi hydrogen explosion which lead to many deaths and dangerous recovery efforts.
Robots have the potential to be useful assistants in situations where humans cannot safely operate, but despite the imaginings of science fiction, the actual robots of today are not yet robust enough to function in many disaster zones nor capable enough to perform the most basic tasks required to help mitigate a crisis situation. The goal of the DRC is to generate groundbreaking research and development in hardware and software that will enable future robots, in tandem with human counterparts, to perform the most hazardous activities in disaster zones, thus reducing casualties and saving lives.
The robot tasks will simulate potential recovery necessities. They will have to drive a vehicle (Polaris) through a course, successfully dismount/exit the vehicle, open a door and walk through it, turn a valve, use a drill to cut a specified shape out of a wall, perform a surprise manipulation task (use their hands) , walk across uneven terrain, and finally climb a ladder. Sounds easy enough, right? Oh wait I forgot to mention that the users are not allowed to see their robots and must control them through the robot’s vision system only. Oh yea, and during a portion of the competition, communications will be degraded and the robot will have to perform tasks autonomously (without user input or control). And finally, the robot must be untethered while executing these tasks. So if it falls, no one and nothing will be there to catch it. With that said, let’s introduce some of the robots!
The most common robot in this competition is ATLAS, developed by Boston Dynamics, which recently underwent some upgrades. ATLAS is designed to be both strong and agile with 28 hydraulically-actuated degrees of freedom (DOF). For sensing it uses stereo cameras and a laser range finder. THORMANG, developed by ROBOTIS in South Korea, will also make an appearance at the DRC. THORMANG has 30 DOF and uses LIDAR as well as an HD camera for sensing. They also incorporated modular components to enable quick replacements and repairs. HUBO (HUmanoid robROt) will be used by two teams. HUBO was developed at the Korean Institute for Science and Technology (KAIST), and comes equipped with three different cameras to visualize 3D objects. Finally, HRP, a product of KAWADA Industries, Inc in Japan, will also be used by two teams. HRP, the 34-DOF lightweight humanoid, is probably the most futuristic-looking robot out of the bunch.
Fifteen teams competed in the DRC Trials in 2013. The first seven teams listed below are in descending order of how many points they scored during trials.
Team IMHC Robotics
The Institute of Human and Machine Cognition (IHMC) in Pensacola, FL will be using an ATLAS robot which they have named Running Man. They are sponsored by both Amazon and Atlassian. The IMHC focus areas include walking and balancing humanoid robots. The team’s backgrounds range from high school students to post-docs. Here is a video of the Running Man performing some of the DRC tasks.
Tartan Rescue came in third during the 2013 Trials. Their robot, CHIMP (CMU (Carnegie Mellon University) Highly Intelligent Mobile Platform) is built with the ability to drive around on four tracks (like a tank), but can also stand upright to perform manipulation tasks (eg. opening a door, or using tools). Its grippers (robot hands) are three-fingered. It uses laser scanners to generate 3D models of the environment.
Team MIT (Massachusetts Institute of Technology) is also using an ATLAS. Their team consists of students from an array of MIT engineering departments, some of which competed in the last DARPA challenge. Team MIT made the choice to keep all of their software open-source while preparing for the competition.
The Jet Propulsion Laboratory was inspired by simian primates when developing this robot. RoboSimian has four identical limbs to deal with mobility and manipulation. The team chose to favor deliberate and stable operations over reaction and dynamic actions to enable fast and robust operations.
TRACLabs is based in Houston, TX. Their approach is to rely on human operators for high-level decision making and to treat the competition as a system integration of the work already taking place at TRACLabs. Using ATLAS, they are working towards improving autonomy and control, and exploring new methods of human-robot interaction.
CMU and WPI (Worcester Polytechnic Institute) teamed up to form Team WPI-CMU. Using an ATLAS named WARNER and a diverse team of many skill sets, they are ready for the competition. Watch them perform the competition tasks below.
Lockheed Martin Advanced Technology Laboratories partnered with the University of Pennsylvania (UPenn). Together, they developed autonomy and control software to guide their robot, Leo. Leo is an ATLAS with an extended forearm used to increase flexibility while performing manipulation tasks. A focus area of theirs was human-robot collaboration.
Team Aero, from Tokyo, Japan, was on a mission to develop a humanoid with a sleek design. Aero DRC is a four-legged robot with two arms and 2-DOF hands. At the base of each leg is a wheel which enables fluid movement in any direction. This team is sponsored by SEED Solutions.
The National Institute of Industrial Science and Technology (AIST) in Japan will be running through the challenge with an HRP robot (HRP-2Kai to be exact). They are partially sponsored by Japan’s Disaster Response National Project.
Team DRC-HUBO is led by the University of Nevada, Las Vegas. HUBO was developed by the Korean Institute for Science and Technology (KAIST). The robot weights about 176 pounds and can lift up to 44 pounds. HUBO has three fingers on one hand and four on the other to perform varied manipulation tasks. Team DRC-HUBO feels that the hardest task will be safely and quickly getting the robot to exit the car.
Team Grit has designed a quadruped to compete in the DRC. Cog-Burn, as they refer to it, is made up of 34 electric servo motors which enable it to walk and use its hands. Team Grit is made up of undergrads and professors from Colorado.
Team Hector is made up of members of the Department of Computer Science at Technische Universität (TU) Darmstadt in Germany. Humanoid robot, Johnny 05, is lightweight (55 kg), tall (1.47 m), and modular to enable easy and quick repairs. Johnny 05’s hands were designed by Virginia Tech’s THOR and VALOR DRC teams. In addition, DRC Team ViGIR provided high level software for the robot’s control and operator interface. Team Hector will open-source all of their code after the competition.
Team HKU (Hong Kong University) will be competing with an ATLAS robot.
Team HRP2-Tokyo will be competing with an HRP robot.
Researchers at Rainbow Co. and KAIST from South Korea collaborated for the DRC, as well as other Korean-based companies. They focused on bipedal walking, articulated manipulation, and unmanned vehicle driving. They are using a HUBO robot. HuboLab developed the control algorithms, Rainbow Co. worked on improvements for the bot (software and hardware), and RCV integrated computer vision (the perception systems on the robot).
Team NEDO-Hydra is made up of four different universities in Tokyo, Japan. The University of Tokyo developed the robot, Hydra, while the other three universities (Chiba Institute of Technology, Osaka University and Kobe University) developed the computer vision, motion and manipulation controls, and network and teleoperation systems. Hydra has 41 DOF, 30 of which belong to the limbs alone.
Team NEDO-JSK is also made up of members from the University of Tokyo. The JSK lab is funded by NEDO (New Energy and Industrial Technology Development Organization). The robot, JAXON, was developed by the lab and incorporates a water-cooled, high-output power drive and control system.
Team NimbRo Rescue
Team NimbRo Rescue comes from the University of Bonn in Germany. The AIS (Autonomous Intelligent Systems) group conducts research in cognitive research and machine learning. Their lab focuses on laser-based and RGB-D mapping and planning. This team has also successfully competed in RoboCups, soccer competitions and a domestic service robot league. Robot Momaro has a flexible hybrid robot base which can roll on steerable wheels and also lift them individually to climb stairs or move over large obstacles.
ROBOTIS, developer of the THORMANG, is using a THORMANG 2. THORMANG 2 is stronger, faster and more stable than THORMANG 1. It is also modular, compact and energy efficient.
Team SNU (Seoul National University) is made up of teammates from the Dynamic Robotics Systems Lab (DYROS), Machine and Intelligence & Pattern Analysis Lab (MIPAL), and SimLab. They are also using a THORMANG by ROBOTIS. DYROS looked into aspects such as biomechanics, human motion analysis and compliant actuators. MIPAL worked on the perception system (how a robot perceives the world through sensing). SimLab provided the interface which operators use to communicate with the robot.
The Robots and Mechanisms Lab (RoMeLa) from the University of California, Los Angeles, and the General Robotics, Automation, Sensing and Perception Lab (GRASP) Lab of UPenn collaborated to form Team THOR. A similar collaboration was successful during RoboCup competitions, and during the development of a humanoid robot (DARwIn-OP) in the past. RoMeLa worked on developing the robotic platforms, and the GRASP Lab focused on the robot’s software, particularly with regards to perception and autonomy. They customized a THORMANG robot to make it more robust and less fragile.
ESCHER (Electric Series Compliant Humanoid for Emergency Response), developed by Virginia Tech engineering students, was built from scratch. It was designed to be lightweight and efficient. Team VALOR collaborated with Team ViGIR on robotic software development. One focus area was to enable both human control and autonomous planning.
ViGIR stands for Virginia-Germany Interdisciplinary Robotics. Team ViGir named their ATLAS robot Florian, after the patron saint of firefighters. Florian uses grippers (robot hands) developed by Robotiq, which incorporates hand mounted cameras and tactile sensing (which enable a robot to “feel”). They plan on making their code open-source after the competition.
WALK-MAN (Whole Body Adaptive Locomotion and Manipulation) was developed by the Italian Institute of Technology and Pisa University in Italy. The WALK-MAN has stereo vision, a 3D laser scanner, and RGB-D and color cameras. IMU sensors are used on the robot’s head and waist to determine the robot’s orientation. They’ve also included soft covers into the design, which protect the robot on impact.
Those are all of the teams that will be competing today. We hope you join us for some live blogging during the event. May the best bot win!