In a year we will see a humanoid robot run 100 meters Maradona Stadium. And, why not, also the long jump, the high jump and other Olympic disciplines. It sounds like science fiction but it is now a consolidated reality which, with the launch of low-cost robots, will become increasingly widespread. For now, however, it is an entirely Neapolitan project. It was created by the Prisma Lab which has been carrying out robotics, mechatronics and automation projects at the Department of Electrical Engineering and Information Technology of the University of Naples Federico II for almost 40 years now, and is the spearhead of global research.
The teachers are at the helm Bruno Sicilianoone of the world’s leading experts in robotics recently awarded the prestigious Pioneer in Robotics and Automation Award 2004 at the international conference in Yokohama, as well as Luigi Villani And Vincenzo Lippiello. The latter is coordinating the project Unina Robotic Olympic Team of the fastest humanoid robot in the world over 100 meters for the future Robotics Olympics made in Naples.
Two months ago, the DIETI noticeboards of Federico II were plastered with announcements like this: «The Prisma Lab is looking for students interested in the use of deep learning, imitation learning, model predictive control and full body control techniques to train a humanoid robot to different Olympic disciplines, starting from the queen discipline of all the Olympics: the 100m race”.
Many responded to the request and a working group was created to train the robot. «The project was a ploy to bring students from various engineering disciplines closer to Artificial Intelligence applied to highly dynamic systems such as robot. It worked: in less than a month I formed a heterogeneous group of almost forty students” he explains Lippiello who does not hide the fact that he was inspired by another project that combines study and play, namely Unina Corse Racing Team, the Federician automotive team that aims to design, build and test a single-seater for various student races. «We wanted to create a similar situation, so that the kids could develop new skills, which could then be transferred to various sectors».
The challenge is to transfer the ability of human running to the machine through AI learning algorithms, deep learning or imitation learning. «Current robotic walking is quasi-static, that is, the robots walk keeping the center of mass of the body which always falls between the two feet and this means in theory that the robot can stop at any time without falling. We also walk like this, but in a running phase we move our center of mass out of a static condition, so much so that if we were to suddenly stop while running, we would fall.”
The input to the project came from the strong development of humanoid robotics. «In the last two years there has been an incredible boom: Elon Musk anticipated in the media something that was already underway, and many producers have accelerated their production – continues the teacher – We are developing the tender procedure for the purchase of the raw robot, and we hope before the summer or in any case immediately after, to have it Supplied.
In the meantime I have already started the project with the kids who are training on a server for the training of artificial intelligence algorithms based on Nvidia, therefore very high-performance, we launched a small competition between them to start developing simpler basic controls to slowly get closer to controlling the humanoid. This is a bit of the context, my ambition would be to be able to have a first attempt at the classic shot race at the Maradona Stadium in a year’s time.”
The humanoid robot on which Prisma Lab has set its sights on the H1 from Chinese Unitree Robotics costing more than that 100 thousand dollars, purchasable within the Nemesi project on European research funds. «It is one of the first humanoid robots for commercial sale with an open platform.
Basically they give the external envelope with basic routines (just walk, ed.) that we can implement with broad-spectrum control algorithms.” In Europe they will be the only university that is planning the sporting development of a humanoid robot, while in China and the USA there have been some projects but only made with robot parts.
© ALL RIGHTS RESERVED
Read the full article at
The morning
