Jan 31 2011
A team of researchers at the University of Vermont have developed a group of virtual inexperienced ‘baby’ robots. These robots need to perform several activities in their simulated world. The tasks include identifying objects and moving towards them. The robots showcasing poor performance will be removed and the best-performing robots will stay on.
The researchers have determined that shape-altering robots learned to walk more quickly. Also their gait was found to be more stable.
Shape-Shifting Robot
Joshua Bongard, Lead Study Researcher and Professor of Computer Science, has mentioned that they have found that shape-changing robots are easy to breed.
Bongard and his colleagues have also developed a real robot that can slowly transform from a snake-like creature without legs to an erect quadruped. They also developed simulated bodies and brains for the virtual robots. Bongard has mentioned that the brains of their robots are made of few dozen virtual neurons and synapses.
The simulated robots vary from today’s real robots by their shape-changing ability as they walk around their simulated world. After performing 5000 simulations, the researchers determined the ideal robot, which can complete all the specific tasks of their simulated environment.
In order to establish their invention, the researchers developed a physical Lego robot with four legs. A brace was included which would slowly move to an erect position from its abdomen. When the legs are bent by the brace, the controller makes the robot move using legs and also bow its spine. It will wriggle around and then slowly will stand up, resembling a coyote.
The researchers are working on developing rapid shape-changing real robots. These robots could be deployed in construction works and can significantly bring down construction cost. They discussed their innovations online in the Proceedings of the National Academy of Sciences on January 10.
Source: http://www.uvm.edu