Seth at 21:10, 27 March 2009

2009-03-27T21:10:22Z

New page

<bibtex>
@conference{dellaert1994tem,
title={{Toward an evolvable model of development for autonomous agent synthesis}},
author={Dellaert, F. and Beer, R.D.},
booktitle={Artificial Life IV, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems},
year={1994}
}
</bibtex>

This is the 94 paper, apparently the 96 paper is good too. More ambitious. They coevolve a gene regulatory network and body in a developmental framework. Th body here is just a dividing grid, each square of which is a different color for a different gene expressed. The evolution has an aesthetic fitness, they just picked the ones that looked nice instead of working it towards some task. I like that they 'got symmetry for free' by simulating cell division. I still don't quite have my head around hwo they evolved the RBN as a gene reg. network, so I'm taking it a bit on faith.

"Our long term goal is the co-evolution of bodies and control systems for complete autonomous agents" I wonder if we are there yet (by their standards).

Luis says that this paper was one of the roots of a later fad that developed around evolving the flags of different countries.

I wrote "Nagpal or Coore?" in the margin. Wonder what I meant by that.

They emphasize a few times in the paper that a strength of it is biological defensibility. OI don't know how I feel about that. I guess it comes down to whether you are defending it's bio-plausibility to a biologist or an outsider.

Dellaert1994tem - Revision history

Seth at 21:10, 27 March 2009