Interactive evolution of dynamical systems

author = "Karl Sims",

title = "Interactive evolution of dynamical systems",

booktitle = "Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life",

year = "1992",

editor = "Francisco J. Varela and Paul Bourgine",

pages = "171--178",

address = "Paris, France",

month = "11-13 " # dec,

publisher = "MIT Press",

keywords = "genetic algorithms, genetic programming, cellular automata, CA, parallel running, connection machine",

ISBN = "0-262-72019-1",

URL = "

http://www.alife.org/conference/ecal-1991",

URL = "

http://www.karlsims.com/papers/DynamicalSystemsECAL92.pdf",

URL = "

https://mitpress.mit.edu/books/toward-practice-autonomous-systems",

size = "8 pages",

abstract = "This paper describes a novel system for creating virtual creatures that move and behave in simulated three-dimensional physical worlds. The morphologies of creatures and the neural systems for controlling their muscle forces are both generated automatically using genetic algorithms. Different fitness evaluation functions are used to direct simulated evolutions towards specific behaviours such as swimming, walking, jumping, and following.A genetic language is presented that uses nodes and connections as its primitive elements to represent directed graphs, which are used to describe both the morphology and the neural circuitry of these creatures. This genetic language defines a hyperspace containing an indefinite number of possible creatures with behaviors,and when it is searched using optimization techniques, a variety of successful and interesting locomotion strategies emerge, some of which would be difficult to invent or build by design.",

notes = "ECAL-91 Interactive evolution of artistic images. Discusses drawback of using GA to define CA state transition tables. Second half, discrete CA states are replaced by one or more continuous variables, whose initial value and rate of change are controlled by differential equations (which may depend upon the cells state and that of its neighbours). Both initial value and rate of change are given by a lisp s-expressions. 5 different types of mutation. 'estimates of computation times are made, and slow expressions are automatically eliminated before being used.' Crossover not clear: may be different from Koza, exchanging only a single node between expressions rather than subtrees. Also when mated both the s-expression controlling the initial state and the rate of change are crossed over. Some runs use complex (ie i,j) rather than real numbers. Run on connection machine, one virtual processor per cell. 256 by 256 arrays processed at interactive rates. Mutations and crossovers performed in a front end machine. Genotypes evolved (interactively) 'in timescales such as 10 minutes'",