The first application of self-evolving capabilities

Experimental model of the University of Oslo (Norway) has added the element " genes " to create the ability to change the machine itself. By enabling / disabling these genes, the system determines the most efficient configuration for accomplishing specific tasks.

This evolution is equivalent to experiencing 20-30 thousand generations before the system finds the optimal design to solve the problem. Of course, this only lasts for a few seconds compared to the 800-900 thousand years that human beings need to go through with the equivalent generation.

Researchers Kyrre Glette and Jim Torresen of the University of Oslo introduced a self-evolving hardware model. (Image: Ståle Skogstad)

The research team of the University of Oslo began the project 'hardware evolution' say on since 2004 with the first product is a humanoid robot chicken named Henriette, which uses software to automatically learn to go.

At present, evolutionary technology is expected to help solve many problems that programmers can not do, because humans can not anticipate all the problems for the machine. For example, when a robot dispatched to Mars crashed into a pit, it could, by its ability to evolve, find its way out of the hole.

One of the most practical applications of the evolutionary technology that researchers are aiming at is to create robots for the installation of oil pipelines and related equipment in the industry at depths of 2,000 meters, The distance from contact with the robot is almost impossible. Then, the robot with the ability to transform itself will be able to find solutions to problems quickly in seconds without the help of people.

PK