How Californian alone won the space robotics competition from NASA

Space robotics competition from NASA (Space Robotics Challenge, SRC) was held last year. During the competition, virtual humanoid robots "Valkyrie" (Valkyrie) roamed the virtual Mars, trying to fix the virtual objects. Everyone could participate, and since the virtuality of the competition meant no need to use large and expensive robots , everyone could really participate. Of the 93 teams that signed up for the competition, NASA selected 20 finalists based on their preliminary tasks for working with the 3D simulator of the Gazebo robot. Each of the finalists was asked to program the humanoid Valkyrie so that he could complete the equipment repair mission on a simulated Martian base.

The winner of the SRC was the Coordinated Robotics team, which, moreover, was the only team that achieved 100% completion of the task, and received the top prize of $ 125,000, plus a bonus of $ 50,000 for “perfect execution”. However, the “team” can be called a stretch, since Coordinated Robotics consists of only one comrade: Kevin Knoedler. We talked with Kevin about the epic winnings, and consulted with Nate Koenig of Open Robotics , leading the development of Gazebo and helping to organize SRC.

SRC was very similar to VRC (qualification in the DARPA Robotics Challenge, DRC) because all teams competed by running code in the virtual Gazebo environment. “The tasks themselves were to some extent inspired by the film 'Martian',” Nate Koenig, technical director of Open Robotics, told us. “The Valkyrie is on Mars, it prepares everything to create a colony of people, and a dust storm rushes in.” After the storm Vale, you need to fix the satellite communication antenna, repair the solar panels and find a leak from the habitable zone. Here are some excerpts from the competition:
')


“Overall, the competition went smoothly,” says Koenig. - The uniqueness of SRC, in contrast to VRC, was that we require consistent execution of tasks. You get more points for completing more tasks when Valkyrie has not fallen, and it didn’t have to be restarted — that is, the more reliably you walk and control objects, the better your results will be. ”

As in the case of DRC, restrictions on the execution time of tasks were set so that the teams tried to make maximum use of the robot offline. Most of them did. To complicate the task, hard constraints were imposed on the data transmission channel and delays that emulated real remote work with a robot somewhere in space. Koenig explains:

Network latency and channel width restrictions were tougher than in the case of VRC. We wanted to simulate something like a delay signal going to Mars and back, but that would be too much - so we agreed on a delay of 20 seconds. Some tasks were limited to 380 bps bandwidth, and this, in essence, is already killing the TCP version.

People needed to be creative in their task, and we saw various unique solutions. One person started the server and the IRC client to transfer information, others used text-based console messages, without visual data — it was like reading the Matrix. One team, Xion, worked completely autonomously: they just wrote the code and clicked the "start" button, as a result of which they managed to accomplish an impressive amount of tasks.

Koenig said that he and his colleagues did not expect any team to be able to complete all the tasks in a row. “But Kevin has denied us,” he adds. “And he turned out to be the only team capable of that.”

Kevin is Kevin Knödler, the full lineup of the Coordinated Robotics team. As Nate said, Kevin was able to complete all the tasks of the SRC without errors and in a row, which is quite surprising. We chatted with him by e-mail to find out how he did it.

We: what is your education, and why did you decide to participate in the SRC on your own?

Kevin: After graduating from MIT, I worked as an engineer and manager at Teradyne. In 2007 I left them to stay at home with a child. Both working there and as a householder, I constantly participated in various competitions - Robot Wars, Battlebots, three competitions on autonomous vehicles from DARPA, and DRC. The SRC seemed difficult and interesting, so I decided to participate in it.

In the autumn of 2016, when the qualifying round began, I was busy training two football teams, and I knew that I would be busy during the final tests at the beginning of 2017, teaching in the program “Odyssey of the Mind”. Usually when working in a team, it is necessary to work together and coordinate in the early stages of the process. Since I knew that I would be busy at key moments, I decided to do everything alone so as not to upset myself and other potential team members. Usually teams work better because the more people, the more creative ideas. On all previous projects I worked with teams.

How much does your strategy rely on autonomy?

I started to develop, assuming that the time delay will be maximum, and the robot will have to perform short tasks on its own. And even without elaboration, the 20-second delay was not the main problem, since the entire competition was given hours. My code was not as perfect and reliable as I would like, so I focused on what I would plan and execute. It was mainly controlled autonomous work with human perception.

You sent us your video from one of the runs (below). Can you comment on it in detail?



The video in the third person perspective shows a robot performing three tasks. The first is to turn the knobs to align the antenna. On the second, the robot takes the solar panel out of the trailer, puts it on the table and connects the cable. In the third, you need to climb the stairs, open the door of the house, find the leak with the tool, and then fix it with the second tool. Most of all I liked the leak finding stage. It was necessary to examine a large area, part of which was blocked, and each time it was interesting to look for a leak.

The leak was found by a robot, a conductive tool up and down with the help of body rotation to minimize the number of steps. The robot, searching for a leak, noted parts of the territory as uninspected, free of leaks, or containing a leak. This information was shown to the operator through an interactive Rviz marker to make it easier for us to see what has already been examined and how the leak was found.

What was the most difficult?

I would say that the most difficult thing was to use and manipulate tools. It was qualitative to take the tool and force the robot to use it, as if it was its continuation, it was difficult to do it every time with the same efficiency. In Gazebo, I created a script where the robot started working next to the tools, and nothing bothered him. This made it possible to test the taking of instruments from different initial positions, followed by placing them on the table.

An interesting fact from the competition: sometimes real equipment gets stuck and you need to make an effort to make it move. The simulation robot behaved the same way. [Open Robotics called this an “interesting emerged behavior,” since it was not specifically introduced into the program]. The robot's thumb could get stuck and not respond to commands. It happened to me at the competition the third run of tasks. But, as in real life, it helped me that I pressed my finger on the table, he stopped sticking and started moving again.

What is simpler in simulations than in reality?

The simulation is easier. Not drastically, but 90% of the problems in the simulation can be solved. Mostly easier because of the reliability of iron - in the simulation it does not break, as often happens with the present. You can also engage in more risky experiments. If a humanoid robot falls in Gazebo, it is not necessary to spend $ 100,000 for a few weeks to repair it. Another advantage - one person can run multiple tests at once. With real robots, several people run one test.

If NASA placed a real Valkyrie at the Mars emulation range and asked you to do the same tasks, what would you do?

The robot must be able to complete all the tasks after the initial checks and correcting the differences between the simulation and the hardware. My approach was multi-level, I could always return to the lower level of control if the main one could not cope. There is always enough difference between a simulation and a real iron in order to create a need for some kind of adaptation for successful development. But after the tests and adaptations, I think everything would work out!

After participating in DRC and SRC, how would you evaluate the potential of humanoid robots in real relief in areas of natural disasters or exploring planets?

After DRC and SRC, we approached the use of humanoid robots in disaster areas and in the exploration of planets. The main difficulties on Earth are reliable iron, capable of handling falls, and working in difficult conditions (the ability to creep under obstacles, the environment with obstacles and lack of free space, a situation where a hand is required for support, etc.). In space, everything is the same plus long distances, which require more perception and autonomy from the robot.


Unsuccessful moments of competitions