Robots respond to punishment at Staffordshire

Robots of the future which can think for themselves and behave more like humans are a step closer to becoming reality thanks to research being carried out at Staffordshire University.

Alex Kleiner, a German postgraduate student studying for an MSc in Computer Science, has developed a software package that uses a system of punishments and rewards to teach a miniature robot to explore its surroundings without bumping into obstacles.

Whenever the five centimetre robot avoids an obstacle in its path, it is rewarded with an electronic pat on the back, a 0 is entered into is databank. If it fails to avoid a collision it is punished with a -1.

Alex explained: Traditional artificial intelligence research has relied on programming a robot to respond to certain situations in a certain way. But, in the real world so many different things can happen, it would be impossible to give a robot instructions to cope with an infinite amount of possibilities. The human brain, however, makes generalisations so we do not have to learn how to react to every eventuality, and the software I am developing tries to teach the robot the same technique.

The early results of Alexís work are proving promising, and throwing up some surprising results. The more tests that are carried out, the more the robot mimics human behaviour.

Like a human, the robot does not learn things first time around. But as it explores its surroundings, the number of collisions decreases, added Alex, who completed a diploma in Computing at college in his home town of Constance in southern Germany.

Alexís work was revealed at the second International Conference on Enterprise Information Systems, hosted by Staffordshire University.

For more information, email Alex Kleiner direct on a.kleiner@staffs.ac.uk

Web site: www.staffs.ac.uk