Autonomous robots may help sweep away 'space junk' from low Earth orbit

May 23, 2021
Robotic arms are the best method to collect human-made space debris. The NanoRacks-Remove Debris Satellite, above, was launched in 2018 from the International Space Station. (NASA/JSC/Drew Feustel)

Robotic arms are the best method to collect human-made space debris. The NanoRacks-Remove Debris Satellite, above, was launched in 2018 from the International Space Station. (NASA/JSC/Drew Feustel)

Low Earth orbit, the area of space where many of our newest, fastest artificial satellites are being deployed, is full of floating trash that cannot reliably be collected, threatening the safety of artificial satellites and the International Space Station. But now a group of researchers has proposed a system that minimizes the challenges of operating robots in space, making it easier for robotic arms attached to spacecraft to collect debris and declutter the cosmos.

In a paper published May 4 in Frontiers in Robotics and AI, researchers from the U.K. and Germany outlined a new, more efficient system for operating robotic arms in space. It marks the first use of imitation learning, or programming by demonstrations, for planning the trajectory of a robot arm on a free-floating spacecraft. 

The team's innovative approach requires less energy and costs less than existing methods, and it can be used to collect debris in space. It comes at a critical time to confront space junk, as the build-up of debris in low Earth orbit is rapidly reducing its capacity to sustain long-term space operations. Physicists are even attempting to expand into the more dangerous medium Earth orbit to find more space. 

R.B. Ashith Shyam, the first author of the paper, led this project as a postdoctoral researcher at the University of Surrey. He told The Academic Times that space robotics is a fast-growing field, as more and more private companies such as SpaceX join governmental space agencies to steer the future of space research. "The revenue from the space sector is projected to reach $1 trillion USD by 2030 and [there is] great opportunity to tap this huge potential," he said. "My work tries to bridge the gap between Earth-based and space robots."

In the paper, the authors explained that robots are hindered in outer space for many reasons. One is communication latency — it takes several minutes for a signal to reach space and get feedback. Power sources are also lacking in space, and robots need to meet extreme safety requirements before being deployed, because things damaged in space cannot be easily or cheaply repaired. Space robots are currently controlled from ground stations or tele-operated remotely by a person from Earth, but the industry is working to make space robots more autonomous so they can independently perform tasks and address challenges.

Robotic arms are the best method to collect human-made space debris, according to the paper, because their design is versatile. In the past, agencies have recommended that debris be collected at the end of space missions. And various other collection methods have been proposed, ranging from harpoons, nets, tentacles and slingshots to an electrical current that would slow down the speed of debris, causing the celestial trash to fall closer to Earth and burn up. 

Shyam and his colleagues think robot arms are the best solution because they can be used in other application areas, such as on-orbit servicing and assembly and autonomous rendezvous and docking. But today's robotic arms that are attached to spacecraft must constantly adjust for orientation disturbances caused by the arm moving to collect debris. Those adjustments consume a significant amount of energy.

Ideally, space-based robotic arms should be able to operate with minimal disturbances to the spacecraft, and they should consume less power, the authors said. "We have introduced a procedure to the space-robotics community [in which] a robot can be trained on Earth, providing it with the capability to make intelligent and autonomous decisions in space," said Shyam. "This is a small step towards fully autonomous space robots. The proposed method is computationally inexpensive and is ideal for space, as the only power source available is the sun."

The researchers relied on imitation learning to train a robot to autonomously perform tasks in space. Shyam said this technique can be compared to how an educator might teach a skill to a child: First the educator demonstrates, and then the child attempts to imitate their actions, gradually learning the skill. Similarly, the robot can be trained while on Earth and remember the actions it learns before being deployed to space to perform those actions in a new setting.

The imitation learning approach is used for the robot's trajectory planning. All learning is carried out offline, which means that minimal power is required for computations after the robot is deployed.

This system is designed to be used by a robot arm with seven degrees of freedom, which are the number of movable joints on the arm; a human arm also has seven degrees of freedom. "One of the main reasons why such an arm is proposed is because it can be used for multiple tasks, for instance, assembly, servicing and capture," Shyam said. "This eliminates the need for role-specific robot arms. The arm design is inspired from the human arm as it caters to a wide variety of tasks."

Shyam and his co-authors used simulations to prove their concept in this study, generating trajectories with the aid of an optimal control algorithm. The next steps for the project would be to conduct real experiments under simulated space conditions of a robot capturing debris.

"This work hopefully will contribute to the ongoing efforts to make space clutter-free," Shyam said.

The study, "Autonomous Robots for Space: Trajectory Learning and Adaptation Using Imitation," published May 4 in Frontiers in Robotics and AI, was authored by R. B. Ashith Shyam, Zhou Hao, Umberto Montanaro, Shilp Dixit, Arunkumar Rathinam, Yang Gao and Saber Fallah, the University of Surrey; and Gerhard Neumann, Karlsruhe Institute of Technology.

Saving
We use cookies to improve your experience on our site and to show you relevant advertising.