Evolved virtual creatures (1994)
First virtual evolution, advanced virtual intelligence
1) Directed by by Carl Sims (Panspermia (1990).
2) This narrated computer animation shows results from a research project involving simulated Darwinian evolutions of virtual block creatures. A population of several hundred creatures is created within a supercomputer, and each creature is tested for their ability to perform a given task, such the ability to swim in a simulated water environment. The successful survive, and their virtual genes containing coded instructions for their growth, are copied, combined, and mutated to make offspring for a new population. The new creatures are again tested, and some may be improvements on their parents. As this cycle of variation and selection continues, creatures with more and more successful behaviors can emerge.
3) Generated autonomous three-dimensional virtual creatures without requiring cumbersome user specifications, design efforts, or knowledge of algorithmic details.
4) The creatures shown are results the final products from many independent simulations in which they were selected for swimming, walking, jumping, following, and competing for control of a green cube.
Results: evolutions were performed for each of the behavior selection methods. A population of interbreeding creatures often converges toward homogeneity, but each separately run evolution can produce completely different locomotion strategies that satisfy the requested behavior. For this reason, many separate evolutions were performed, each for 50 to 100 generations, and the most successful
creatures of each evolution were inspected. A variety of more complex strategies also emerged from some evolutions. A few creatures pulled themselves through the water with specialized sculling appendages. Some used two symmetrical flippers or even large numbers of similar flippers to propel themselves, and several multi-segmented watersnake creatures evolved that wind through the water with sinusoidal motions, direction of future work might be to adjust the genetic language of possible creatures to describe only those that could actually be built as real robots. The virtual robots that can best perform a given task in simulation would then be assembled, and would hopefully also perform well in reality. Much work could be done to dress up these virtual creatures to give them different shapes and improved rendered looks. Flexible skin could surround or be controlled by the rigid components. Various materials could be added such as scales, hair, fur, eyes, or tentacles, and they might flow or bounce using simple local dynamic simulations, even if they did not influence the overall dynamics.
5) Artificial evolution permits the generation of complicated virtual systems without requiring design, and the use of unbounded genetic languages allows evolving systems to increase in complexity beyond our understanding.