DARPA Robotics Challenge 2015
If you couldn’t make the event here’s my blog ……
The biggest robotics event ever, was held in Pomona, California on June 5-6
Background
The winner was worth $2M, which is a large amount by any standards (2nd place got $1M and 3rd place got $500k) and the challenge was to navigate a series of obstacles designed around a disaster scenario. The robot had to drive a car, get out and open a door. Then go into a room and close a valve. Then cut through a wall using a cordless drill and retrieve and object before walking over rubble and up some stairs. The teams knew about the tasks to be done but not the detail about them. They also had to overcome a surprise task (turned out to be switching on and off, pressing a button and to remove and replace a cable plug.
Typically, most current robots are limited to pristine and precisely controlled settings like factories and research labs, employed in simple, repetitive tasks. These kinds of robots cannot deal well with unpredictability and unstructured environments.
What makes this particular event extremely difficult is that the robots aren’t remote controlled. They are semi-autonomous (in this case “supervised” or “task-level” autonomy). This means that a human operator could issue a robot a command like “Open the door” and the robot would need to be able to complete that task by itself, taking into account the sensing and motions involved in identifying a door handle, applying the right force, and appropriately manoeuvring its limbs.
There were 23 teams in the finals from the United States (12), Japan (4), South Korea (3), Germany (3), Italy (1). There were none from the UK. All had to pre-qualify meaning these were the best of the best in the world. There were two major platforms (Atlas and Thor) both based on two-legged humanoids. Then there were some other slightly different platforms based on more legs and/or wheels.
DARPA were not expecting any of the robots to be even close to matching the capabilities of a human. Even though they are incredible mechanical systems controlled by highly sophisticated software, it’s just the start of a long process of improvement.
Challenge
The challenge itself was impressively set up in an old art deco racetrack with stands for spectators numbering perhaps a 1,000 people. They had four courses set up where teams concurrently tried to compete against each other. Each course had a huge video screen so you could see what was going on.
I was expecting the challenge to prove rather slow and that was certainly the case. It took 10-15 minutes just to walk through an open door (worth 1 point) to polite applause from the crowd. There were plenty of falls (the robots had to get themselves back up) and if they needed help they had to go back to the start of that particular part of the challenge.
Taking first place and the $2 million in prize money was Team Kaist (South Korea), and its robot DRC-Hubo. Coming in second and taking home $1 million was Team IHMC Robotics (Pensacola, Florida), and its robot Running Man. The third place finisher, earning the $500,000 prize was Tartan Rescue (Pittsburgh, CMU), and its robot CHIMP. All three teams got full scores for completing the tasks but Team Kaist won because they completed the course in the fastest time (44 minutes). Bear in mind a human could have done the same course in about 5 minutes, it shows just how far we have to go! However, the fact that 19 robots managed to complete the course was a massive improvement on two years ago (where none of them managed it) shows the pace of change possible given the right incentive.
Expo
Surrounding the competition was an exhibition of over 70 exhibitors. They showcased a huge range of robotics and there were regular demonstrations throughout (including aerial vehicles). Most exhibitors had robots that attendees could play with or they were whizzing around the grounds.
Robotics Database
We attended the event to ensure we were as up to date as possible with the latest robotics technologies. It was great to see that we had all the exhibitors (and their robots) already included meaning we are as up-to-date as possible.
Summary
It was an event well worth attending. There were big crowds and lots of very interesting technologies on show. Clearly, robots have a long way to go before they take over from humans in carrying out unstructured tasks in uncertain environments but you have to start somewhere.
If you would like more information please contact me.