Cellular Encoding for Interactive Evolutionary Robotics   [CE] [ER]

by

Gruau, F. and Quatramaran, K.

Literature search on Evolutionary ComputationBBase ©1999-2013, Rasmus K. Ursem
     Home · Search · Adv. search · Authors · Login · Add entries   Webmaster
Note to authors: Please submit your bibliography and contact information - online papers are more frequently cited.

Info: 1996
Keywords:genetic algorithms, genetic programming
Abstract:
This work reports experiments in interactive evolutionary robotics. [ER] The goal is to evolve an Artificial Neural Network (ANN) to control [NN] the locomotion of an 8-legged robot. The ANNs are encoded using a cellular developmental process called cellular encoding. [CE] In a previous work similar experiments have been carried on successfully on a simulated robot. They took however around 1 million different ANN evaluations. In this work the fitness is determined on a real robot, and no more than a few hundreds evaluations can be performed. Various ideas were implemented so as to decrease the required number of evaluations from 1 million to 200. First we used cell cloning and link typing. Second we did as many things as possible interactively: interactive problem decomposition, interactive [PD] syntactic constraints, interactive fitness. More precisely: 1- A modular design was chosen where a controller for an individual leg, with a precise neuronal interface was developed. 2- Syntactic constraints were used to promote useful building blocs and impose an 8-fold symmetry. 3- We determine the fitness interactively by hand. We can reward features that would otherwise be very difficult to locate automatically. Interactive evolutionary robotics [ER] turns out to be quite successful, in the first bug-free run a global locomotion controller that is faster than a programmed controller could be evolved.
URL(s):(G)zipped postscript
Other format

Review item:

Mark as doublet (will be reviewed)

Print entry




BibTex:
@TechReport{gruau:1996:ceier,
  author =       "Frederic Gruau and Kameel Quatramaran",
  title =        "Cellular Encoding for Interactive Evolutionary
                 Robotics",
  institution =  "School of Cognitive and Computing Sciences, University
                 of Sussex",
  year =         "1996",
  type =         "Cognitive Science Research Paper",
  number =       "425",
  address =      "Falmer, Brighton, Sussex, UK",
  month =        jul,
  keywords =     "genetic algorithms, genetic programming",
  URL =          "ftp://ftp.cogs.susx.ac.uk/pub/reports/csrp/csrp425.ps.Z",
  url_2 =        "http://www.cogs.susx.ac.uk/cgi-bin/htmlcogsreps?csrp425",
  abstract =     "This work reports experiments in interactive
                 evolutionary robotics. The goal is to evolve an
                 Artificial Neural Network (ANN) to control the
                 locomotion of an 8-legged robot. The ANNs are encoded
                 using a cellular developmental process called cellular
                 encoding. In a previous work similar experiments have
                 been carried on successfully on a simulated robot. They
                 took however around 1 million different ANN
                 evaluations. In this work the fitness is determined on
                 a real robot, and no more than a few hundreds
                 evaluations can be performed. Various ideas were
                 implemented so as to decrease the required number of
                 evaluations from 1 million to 200. First we used cell
                 cloning and link typing. Second we did as many things
                 as possible interactively: interactive problem
                 decomposition, interactive syntactic constraints,
                 interactive fitness. More precisely: 1- A modular
                 design was chosen where a controller for an individual
                 leg, with a precise neuronal interface was developed.
                 2- Syntactic constraints were used to promote useful
                 building blocs and impose an 8-fold symmetry. 3- We
                 determine the fitness interactively by hand. We can
                 reward features that would otherwise be very difficult
                 to locate automatically. Interactive evolutionary
                 robotics turns out to be quite successful, in the first
                 bug-free run a global locomotion controller that is
                 faster than a programmed controller could be evolved.",
  size =         "23 pages",
}