Robots are an indispensable part of space exploration. At the moment, however, machines like the Mars rover Curiosity are more reminiscent of spartan remote-controlled cars. This is set to change in the future: Human-like robots are also to be used. For example, these robonauts could be controlled from a safe distance in order to carry out dangerous tasks. The MyelinS start-up offers exactly the right software for such a mission.
»Our software offers three essential functions,« says Zied Tayeb, PhD student at the Chair of Cognitive Systems at TUM and, together with Samaher Garbaya, one of the minds behind MyelinS. »The first function is navigation. We use machine learning to let robots map their environment and avoid obstacles.« This is also important for remote-controlled robots. This allows humans to concentrate on the core task of the mission, while the robot – one also speaks of shared control. »The second function is tactile feedback,« explains Tayeb. »Our software can learn to output information from the robot's tactile sensors via vibration motors. For example, if a robot touches rock samples, the human remote control would be able to feel the structure of those samples.«
Curiosity as a Feature
The third feature of MyelinS software is curiosity, which has often been the beginning of great discoveries. »We use machine learning to teach our software what looks 'interesting' and should be highlighted,« says Zied Tayeb. »Depending on which expert trained the algorithm, it could be strange rocks, deformed machine parts, or three-eyed fish.« The MyelinS software is platform-independent and could therefore be used for many different robots.
Intelligent Prostheses as Starting Point
The control of space robots was not the first idea of Zied Tayeb and Samaher Garbayer. Tayeb had developed the open source software package Gumpy, which was designed for brain-computer interfaces. This includes, for example, the basics of the current software function »tactile feedback«. »We wanted to develop software for prostheses that made it possible to feel the artificial limbs,« explains Tayeb. »We developed Gumpy further and added better algorithms and more functions.«
According to Tayeb himself, he spoke to an old mentor about a year ago: »When he heard what we were doing, he said that the major space agencies might be interested in brain-computer interfaces.« It wasn't easy for Garbayer and Tayeb to say goodbye to the prosthesis project, but they didn't want to miss the opportunity to participate in space exploration.
Support by the TUM
MyelinS was supported by TUM Gründungsberatung. In addition, the team used the services of UnternehmerTUM, the centre for innovation and entrepreneurship. Among other things they took part in the Xplore Pre-Incubation Bootcamp. Gordon Cheng, Professor of Cognitive Systems at TUM, also supported the team. In the coming months he and the team also want to formally found their company.
MyelinS was the only European team to participate in the finals of the NASA iTech Cycle II competition. On October 7th and 8th, MyelinS presented the software in Las Cruces, New Mexico. The teams can win mentors who accompany the projects on their way and advise on how to make their products fit for the market.