< IMG SRC = "/Uploads/Blogs/35/0D/IB-FQNFMAAQ3_DF4A499.jpg" Alt = "This Chinese Robot is extremely skillfully cope with the biggest problem of robots"/> ~ ~ ~ < P > One of the biggest problems of humanoid robots is the ability to raise after falling. That is why we see on various demonstration rollers as robots try to push and throw. Researchers from China have created a work-humanoid that seems to have mastered the ability to get up, and does it in an interesting and slightly serf style.
~ < p >< strng > details
< P > Researchers shared a video that shows how two -legged humanoids stand up, lying on the back, sitting near the wall, lying on the sofa and throwing in a chair. Researchers have also checked the ability to correct on different areas and slopes, including Kam & Rsquo; Jan road, glass slope and relying on a tree.
< p >The work was even able to get up when the developers interfered with him. In each script, you can see how the robot adapts to the environment and successfully gets up.
< p >< iframe width = "560" Height = "314" SRC = "//www.youtube.com/embed/yruh-3cfwe4" Alowfullscreen = "ALLOWFULLSCREN" >0 ~/IFrame >0 ~/P > < P > The resistance of the work is caused by the system called “Humanoid Standing-Up Control” (Host). Scientists have achieved this through supporting, type of machine learning, where the agent (in this case, the Host structure) tries to complete the task of attempts and errors. In essence, the robot performs a certain action, and if this action leads to a positive result, he is sent a signal of remuneration that encourages him to perform this action again next time when he finds himself in a similar state.
~ ~ < P > Framework Host initially studied with the ISAAC GYM simulator, the physics simulation environment developed by NVIDIA. After the infrastructure has been sufficiently trained in simulation, it was launched at the UNITREE G1 humanoid work for experimental testing, the results of which are shown in the video. Scientists say:< blockquote > < P > The results of the experiment with the Humanoid Robot UnitREe G1 demonstrate smooth, stable and reliable movements of upset in different real -world scenarios. Looking forward, this work paves the way to integrate systems of supply into existing humanoid systems with the potential of expanding their real use.