ROBOTS CAN LEARN TO EVOLVE WITHOUT HUMAN HELP

Robotics Project Could Provide Next Generation "Pro-Active" Production Robots

Researchers have created a robotic system that can learn from its mistakes and improve its performance. Essentially they believe they have witnessed “the process of evolution by natural selection in robots”. The ‘mother’ robot can design, build and test its own ‘offspring’, and then learn from the results to improve on the design of the next generation, without the support of computer simulation or human assistance. The ‘baby’ robots are plastic cubes that are powered by a simple internal motor. These are assembled by the ‘mother’ robot arm which sticks them together in different arrangements. The mother robot evaluates each of its babies performance on how far it can move, and with no human interference, revises the design so the next one it constructs is able to move further. The mother robot went on to construct ten generations of ‘offspring’. The final generation was able to move twice as far as the first before it ran out of power.

According to lead researcher Dr Fumiya Iida of Cambridge University one of the aims of the research was to discover new insights into how living things evolve over time. He told the BBC, “One of the big questions in biology is how intelligence came about – we’re using robotics to explore this mystery. We think of robots as performing repetitive tasks, and they’re typically designed for mass production instead of mass customisation, but we want to see robots that are capable of innovation and creativity. His colleague Andre Rosendo revealed that another aim of the project was to develop robots that can learn to get better and adapt to new situations. “You can imagine cars being built in factories and the robot looking for defects in the car and fixing them by itself.”