This air gap, thinner than a strand of hair and so hardly visible to the human eye, allows the platforms to move across the floor without any friction, reproducing the state of weightless free-floating in two dimensions.
“The algorithm that makes Olympus move is trained using reinforcement learning – a machine learning method that works on the basis of trial and error. This means the robot controls its orientation autonomously,” Jørgen adds.
When the platform rotates to face one direction, the robot tries to right itself with a swimming-like motion, using a technique it has identifies as the best one during simulations. “This configuration, with Olympus attached to one of ORBIT’s floating platforms, allows us to test the legs’ full range of motion. During one of the testing setups, Olympus was even able to move from wall to wall, reorienting itself after each jump to always land on all four ‘feet’.”
Jules Noirant of ESA’s Orbital Robotics Laboratory comments: “Jørgen’s stay at the ORL highlights the versatility of our testing facilities and their ability to support robotic exploration. Our activities range from locomotion controllers in a planetary environment to jumps and stabilisation techniques in microgravity. We are always pleased to host PhD candidates as visiting researchers to validate their work in a relevant environment, creating a valuable outcome for their thesis.”