Embodied Artificial Evolution Projects

Artificial Intelligence Laboratory

E-mail: [dbisig|bongard|dravid|eggen|llicht|thomas]@ifi.unizh.ch
Fax: +41-1-635 68 09

Research Summary

All forms of natural intelligence are the result of a process of evolution: they have evolved as physically embodied systems. In spite of this, traditional approaches for evolving artificial agents and/or robots tend to focus on isolated parts of an agent, such as the neural control structure. In our approach, referred to as Embodied Artificial Evolution, artificial evolution must exploit the interplay between morphology, neural control, physical materials and the environment, as in natural evolution. For an example, see the evolution of an artificial compound eye.

Much of our work is focused on including into the evolutionary process mechanisms which seem to enhance the exploitation of these interdependencies: mechanisms which we are currently investigating include development, self-organization, and the evolution of evolvability, in which the adaptive potential of evolving systems is itself subject to selection pressure.

Another goal of this group is to find new ways to automatically explore "life as it could be": in other words, generating diversity of artificial life forms, either as simulated agents or as real-world robots. By using EAE, we minimize designer bias, and also produce a "track record" of the incremental improvement in the design process. It thus may be possible for a human designer to subsequently abstract design methods for producing intelligent robots. In addition, EAE may prove to be the only feasible method for generating and controlling robots which achieve increasingly complex tasks, especially when they come in huge numbers, as in applied nanotechnology.
Projects and People

The Evolution of Non-Trivial Behaviour
Raja Dravid
The Artificial Ontogeny Project
Josh Bongard
Artificial Growth Mechanisms
Dale Thomas
Evolving Sensor Morphologies using Adaptive Hardware
Lukas Lichtensteiger
Reconfigurable Modular Robots
Lukas Lichtensteiger
Artificial Morphogenesis
Daniel Bisig
The Artificial Evolutionary System (AES)
Peter Eggenberger
Selected Publications

Lichtensteiger, L., Eggenberger, P. (1999) Evolving the Morphology of a Compound Eye on a Robot in Proceedings of the Third European Workshop on Advanced Mobile Robots (Eurobot '99), 6-8 September 1999. pp. 127-134.

Eggenberger, P., Dravid, R. (1999) An Evolutionary Approach to Pattern Formation Mechanisms on Lepidopteran Wings, in Congress on Evolutionary Computation 1999, Vol. 1, pp. 470-473.

Bongard, J. C., Paul, C. (2000) Investigating Morphological Symmetry and Locomotive Efficiency using Virtual Embodied Evolution in J.-A. Meyer et al (eds.), From Animals to Animats: The Sixth International Conference on the Simulation of Adaptive Behaviour. pp. 420-429.

Bongard, J. C. & R. Pfeifer (2001) Repeated Structure and Dissociation of Genotypic and Phenotypic Complexity in Artificial Ontogeny in Spector, L. et al (eds.), Proceedings of The Genetic and Evolutionary Computation Conference, GECCO-2001. San Francisco, CA: Morgan Kaufmann publishers, pp. 829-836.
[Go to AILab] Go to AILab main page Go to IFI main page [Go to IFI]