Now Robot Can Learn To Walk Without Prior Knowledge

There is another achievement in the field of robotics and biologists that the team of USC researchers at the USC Viterbi School of Engineering believes that they have created an AI controlled robotic limbs driven by animal like tendons. It can even be tripped up then recover between the time of the next footfalls.

A professor of Biomedical Engineering Francisco J. Valero-Cuevas and Professor Biokinesiology & Physical Therapy at USC in a project with USC doctoral student Ali Marjaninejad and two other doctoral students -- Dario Urbina-Melendez and Brian Cohn have developed a bio inspired algorithm. This algorithm can even walk itself after only 5 minutes of unstructured play and then take some other tasks without any additional programme.

In the article of Nature Machine Intelligence has explained exciting possibilities for understanding human movements and disabilities, creating responsive prosthetics and robots. This technology can interact with complex and changing environment like space explorations and search and rescue.

A doctoral candidate whose name is Marjaninejad and the research paper’s lead author said this breakthrough is akin to the natural learning that happens in babies Marjaninejad explain; the robot was first allowed to understand its environment in a mechanism of free play. The paper’s author says that, unlike most current work, their robots learn by doing and without any prior on parallel computer simulations to guide learning.

Marjaninejad also added the importance of the programme because they have predicted and code for multiple scenario but not for all scenario thus preprogrammed robots are inevitably prone to failure.

Through the process of discovering their body and environment the robot limbs are designed at Valero Cuevas' lab at USC. USE their unique experience to develop the pattern that works enough for them. Producing the robots with personalized the movements.

By: Lakshender S Angras

Content: https://www.sciencedaily.com/releases/2019/03/190311125138.htm