cedric sileo : ingénieur, intelligence artificielle, docteur en sciences

lundi | 24 juillet 2017

R Daneel se relève seul

R DaneelA humanoid robot with an exceptionally nimble knack for getting back on its feet after a fall has been developed by researchers in Japan.

Named R Daneel, the robot kicks up its legs and rolls back onto its shoulders to gain the momentum it needs to rock up onto its feet and into a crouching position. This might be fairly easy for a human to do, but for the 60-kilogram bot, it requires a relaxed attitude to body control.

 

 

"The robot is not controlled all the time through a predefined trajectory - as is typically done in robotics," says Max Lungarella, who works at the lab where R Daneel was developed, at University of Tokyo in Japan, but who is not directly involved with the project. R Daneel - the R is for Robot - was developed by Yasuo Kuniyoshi and takes its name from a humanoid robot that appears in several books by the late science fiction author Isaac Asimov.

The Japanese bot boasts a sophisticated series of sensors, including gyroscopes, accelerometers, and torque sensors. But unlike most humanoid bots, it is designed to embrace a lack of constant control and instead follow the trajectory determined by the weight and shape of its body during the rocking motion, until it lands square back on its feet.

Flexibility and grace

"At that point control goes back to the robot's brain which is ultimately responsible for integrating the information coming from its various sensors," Lungarella explains.

A video clip of R Daneel in action rolling itself back to its feet can be seen in a video on the researchers' website (13MB Mpeg).

The research project is aimed at exploring more flexible - and graceful - ways for robots to interact with the world around them. "The main idea behind the design of the robot is the exploitation of body dynamics," Lungarella told New Scientist.

The same blend of control and flexibility used in standing up could also be applied to other robot tasks, Lungarella believes. "All kinds of tasks - particularly dynamics-based ones - can be addressed with our framework. We are currently looking at jumping, rolling, walking, trotting, swinging, reaching and grasping."

Dim lights Embed Embed this video on your site

(Sources New Scientist)

 

chef de projet web technologies intelligence information formation chef projet artificielle doctorat informatique maîtrise développement thèse université systèmes recherche société temps base machines machine problèmes connaissances problème techniques exemple système règles internet définition