ROBOTRONIKA - hypermatic:automagic
19 - 23 June 1998 Museumsquartier, Vienna/Austria

Prof. Inman Harvey, University of Sussex, Centre for Computational Neuroscience and Robotics, UK

"Artificial Life through Evolution"

Building artificial animals like robots is just incredibly difficult, and we have failed to produce anything as impressive as the very simplest natural creature. These natural organisms were designed through natural Darwinian evolution - so let's try and learn from this.

We use a population of robots, with their `nervous systems' specified by strings of `artificial DNA', and test each one in turn on the job we want it to do. Initially, with random DNA, they are all pretty useless, but those that are slightly better than average get to survive, and have children who inherit their parent's DNA - robot sex, except this bit is done very discreetly inside a computer.

The next generation is slightly better than the first, we get to test them all and once again `survival of the fittest' decides which robots get to have children and grandchildren in succeeding generations. Continue for enough generations, and with a bit of luck you get robots that behave as you wanted - though no human did the actual design of the control system, and very often we don't have a clue *how* they work! Maggie the 8-legged walking robot will demonstrate that it can be done.

Inman Harvey is a Research Fellow in Evolutionary Robotics at the CCNR, University of Sussex, having written the first Doctoral thesis on this area. The group he works with are now world leaders in applying artificial evolution to all sorts of complex design tasks. He shares a lab with 10,023 other organisms, some more sensible than others: 14 humans, 9 robots and 10,000 ants.