From alife@COGNET.UCLA.EDU Mon Apr 5 233732 1993 Return-Path +lt;alife@COGNET.UCLA.EDU+gt;

---
Master Index Current Directory Index Go to SkepticTank Go to Human Rights activist Keith Henson Go to Scientology cult

Skeptic Tank!

From alife@COGNET.UCLA.EDU Mon Apr 5 23:37:32 1993 Return-Path: Received: from Regulus.COGNET.UCLA.EDU by world.std.com (5.65c/Spike-2.0) id AA19466; Mon, 5 Apr 1993 23:37:14 -0400 Received: by regulus.cognet.ucla.edu (Sendmail 5.61c+YP/3.20-COG) id AA01031; Fri, 26 Mar 93 18:11:53 -0800 Date: Fri, 26 Mar 93 18:11:53 -0800 From: alife@COGNET.UCLA.EDU Message-Id: <9303270211.AA01031@regulus.cognet.ucla.edu> To: alife@COGNET.UCLA.EDU Subject: Alife Digest Volume #097 Status: R Alife Digest, Number 097 Friday, March 26th 1993 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ Artificial Life Distribution List ~ ~ ~ ~ All submissions for distribution to: alife@cognet.ucla.edu ~ ~ All list subscriber additions, deletions, or administrative details to: ~ ~ alife-request@cognet.ucla.edu ~ ~ All software, tech reports to Alife depository through ~ ~ anonymous ftp at ftp.cognet.ucla.edu in ~ftp/pub/alife (128.97.50.19) ~ ~ ~ ~ List maintainers: Liane Gabora and Rob Collins ~ ~ Artificial Life Research Group, UCLA ~ ~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Today's Topics: Calendar of Alife-related Events Paper Available Emergent Robot Behavior using Classifier Systems Nonlinear Science Preprint Bulletin Board Announcing the hodge-podge machine... Animal Behaviour Software Available Announcement of Availability of CA software Evolutionary Robotics - Tech. Reports ---------------------------------------------------------------------- Date: Fri, 26 Mar 93 18:05:36 -0800 From: liane@CS.UCLA.EDU (Liane Gabora) Subject: Calendar of Alife-related Events ********************************************************************** Conf on Fuzzy Systems, San Francisco CA Mar 28-Apr 1, 1993 v79 AI and Simulation of Behaviour Conf, Birmingham UK Mar 29-Apr 2, 1993 v75 Intnl Conf on Neural Nets and GAs, Innsbruck, Austria Apr 13-16, 1993 v80 BEAM Robot Olympics, Toronto Canada Apr 22-25, 1993 v81 Workshop On Computational Neurosciences, Austin, TX May 14-15, 1993 v94 European Conf on ALife, Brussels May 24-26, 1993 v82 Intnl Workshop Neural Networks, Barcelona Spain June 9-11, 1993 v76 World Congress on Neural Networks, Portland, OR July 11-15, 1993 v95 Intelligent Systems for Molecular Biology, Washington July 7-9, 1993 v84 Fifth Intnl Conf on GAs, Urbana-Champaign IL July 17-22, 1993 v80 Dynamically Interacting Robots Workshop Late Aug, 1993 v91 Congress on Medical Informatics, Sao Paulo, Brazil Sept 9-14, 1995 v91 (Send announcements of other activities to alife@cognet.ucla.edu) ********************************************************************** ------------------------------ Date: Wed, 17 Mar 93 15:55:31 MST From: mm@santafe.edu Subject: Paper Available The following paper is available by public ftp. Revisiting the Edge of Chaos: Evolving Cellular Automata to Perform Computations Melanie Mitchell Peter T. Hraber James P. Crutchfield Santa Fe Institute Santa Fe Institute University of California, Berkeley Santa Fe Institute Working Paper 93-03-014 Abstract We present results from an experiment similar to one performed by Packard (1988), in which a genetic algorithm is used to evolve cellular automata (CA) to perform a particular computational task. Packard examined the frequency of evolved CA rules as a function of Langton's lambda parameter (Langton, 1990), and interpreted the results of his experiment as giving evidence for the following two hypotheses: (1) CA rules able to perform complex computations are most likely to be found near ``critical'' lambda values, which have been claimed to correlate with a phase transition between ordered and chaotic behavioral regimes for CA; (2) When CA rules are evolved to perform a complex computation, evolution will tend to select rules with lambda values close to the critical values. Our experiment produced very different results, and we suggest that the interpretation of the original results is not correct. We also review and discuss issues related to lambda, dynamical-behavior classes, and computation in CA. The main constructive results of our study are identifying the emergence and competition of computational strategies and analyzing the central role of symmetries in an evolutionary system. In particular, we demonstrate how symmetry breaking can impede the evolution toward higher computational capability. To obtain an electronic copy: ftp santafe.edu login: anonymous password: cd /pub/Users/mm binary get rev-edge.ps.Z quit Then at your system: uncompress rev-edge.ps.Z lpr -P rev-edge.ps To obtain a hard copy, send a request to mm@santafe.edu. ------------------------------ From: A.Fraser@eee.salford.ac.uk Date: 19 Mar 93 14:46 Subject: Emergent Robot Behavior using Classifier Systems At Salford University, within the Mobile Robots Research Group, we are attempting to develop classifier systems which will produce emergent behaviours for the control of mini-mobots. While some work has already been done with neural networks (see Randall Beers , Intelligence as Adapative Behaviour: An Investigation into Computational Neuroethology) we feel the abilty to be able to evolve new behaviours is necessary. In the near future we can envisage complex tasks being asked of a number of mobile robots which will then use genetic algorithms to evolve, from maybe a base of initial behaviours or a completely blank behavioural base, the behaviours needed to complete the task. Our initial forays into classifier systems have been to try and re-implement UCLA's Santa Fe (or John Muir) trail where an artificial insect attempts to follow a pheremone trail with progressively more complex turns (see D.Jefferson et al Artificial Life II and J.Koza Artificail Life II). We are also looking into learning in classifier systems which will be applied in such a way that a robotic device can adapt on-line to a dynamic environment. While implementing such systems we have come across problems not discussed in papers such as: should a classifier, which is created by a random function, be able to have a large number of possible conditions linked by logical operators and if so how? If anyone has tried to implement a classifier (preferably, but not necessarily in C++) and has any advice to give then please e-mail us. Also if anyone is working in the field of evolutionary robotics (see Alife Digest 94) we would be pleased to hear from you. Thanks in advance.... Adam P.Fraser John R. Rush ============================================================================= || ||\\ /|| ||\\ ||\\ //\\ || A.P.Fraser, || || || \\ / || || \\ || \\ // || Snail:PostGraduate Section || || || / || || // || // // \\\ || Elec & Electronic Dept || || || || ||//\ ||//\ \\ // || University Of Salford || || || || || \\ || \\ \\// || Salford, M5 4WT, England || || ...Mobile Robots Research Group... || Email:A.Fraser@eee.salford.ac.uk || ============================================================================= |||| When we try to pick anything out by itself, we find |||| |||| it hitched to everything else in the universe. - John Muir |||| ============================================================================= ------------------------------ Date: Fri, 19 Mar 93 21:28:36 MST From: erica@huoshan.lanl.gov (Ren Jun-Rui) Subject: Nonlinear Science Preprint Bulletin Board {*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*} {*}+ +{*} {*}+ NONLINEAR SCIENCE +{*} {*}+ PREPRINT BULLETIN BOARD (PBB) +{*} {*}+ +{*} {*}+ Coordinated by the +{*} {*}+ Center for Nonlinear Studies (CNLS) +{*} {*}+ Software developed by Theoretical Division +{*} {*}+ Los Alamos National Laboratory +{*} {*}+ +{*} {*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*} An electronic Nonlinear Science Preprint Bulletin Board (PBB) has been established as of March 5, 1993. The PBB provides a fully automated system for the archiving and distribution of electronic preprints. Preprints are submitted electronically to the system, which assigns them reference numbers, and makes them available to PBB users via a wide variety of conventional means of network access. The simplest such access is via e-mail request, through which remote users may, for example, get help on available commands, obtain the full texts of papers, obtain listings for given periods, and search for author names or keywords. The system allows the original submitter of a paper to incorporate ongoing corrections and addenda (and adds an entry to the daily listings with an author supplied synopsis of any changes). The system also permits anonymous FTP access to papers and macros stored in the data base, as well as access by other rapidly developing network utilities such as WAIS, Gopher, and WorldWideWeb. A cross-referencing feature is available to establish linkages among bulletin boards in different disciplines. All researchers in the nonlinear science community are invited to submit electronic preprints to the Nonlinear Science PBB. For purposes of standardization, users are encouraged, albeit not required, to submit text in TeX or LaTeX format, with figures in uuencoded tar-compressed postscript files. (If necessary, papers may also be submitted with a note indicating that non-electronic figures are to be obtained directly from the authors.) PBB subscribers are notified regularly by e-mail of new submissions. Listings, retrievals, and searches of preprints may be performed at any time without subscribing to the regular mailing service. Usage of the PBB is free of charge. The nonlinear science PBB is divided into the following categories: CATEGORY INTERNET ADDRESS ADVISORS dynamical systems/chaos/ chao-dyn@xyz.lanl.gov Predrag Cvitanovich quantum chaos/topological Mitchell Feigenbaum dynamics/cycle expansions John Guckenheimer turbulence/propagation Ronnie Mainieri Jerrold Marsden Michael Tabor pattern formation/coherent patt-sol@xyz.lanl.gov * Guenter Ahlers structures/integrable Alan Bishop systems/solitons David Campbell Sue Coppersmith Irving Epstein Alan Newell adaptation/interacting adap-org@xyz.lanl.gov Phil Anderson particle systems/self- James Crutchfield organizing systems/ David Griffeath computation theory/ Charles Taylor machine learning Jordan Pollack Daniel Stein computational methods/time comp-gas@xyz.lanl.gov * Gregory Beylkin series analysis/signal Gary Doolen processing/wavelets/lattice James Glimm gases Brosl Hasslacher James M. Hyman George Zweig [all of the above] nlin-sys@xyz.lanl.gov Erica Jen * unconfirmed Users may access and/or subscribe to any subset of the above categories, or to nlin-sys which contains the entire data base; preprints should however be submitted to only a single category since cross-linkage options, as explained in the documentation for any of the categories, eliminate the need for multiple storage. Information on usage of the PBB may be obtained by sending an e-mail message with the subject "help" and a blank message body to any of the above internet addresses; e.g.. To: nlin-sys@xyz.lanl.gov Subject: help Included in the "help" documentation are instructions for the use of other commands: subscribe get put listing cross replace cancel add find distribution published comment typically invoked, as in the case of "help," by sending e-mail to the proper internet address with a subject consisting of the command and requisite arguments, together with a message body where appropriate. To subscribe to the full nlin-sys, for example, send e-mail with a blank message body To: nlin-sys@xyz.lanl.gov Subject: subscribe [your name] replacing [your name] by your full name (spaces and initials allowed) as you wish it to appear in the distribution list. Preprint notifications will then be sent to the e-mail address from which you subscribed. Any of the categories [chao-dyn patt-sol adap-org comp-gas] may be substituted for nlin-sys if you wish to subscribe only to that subset. PBBs currently in operation in other disciplines, utilizing the same software and to which cross-linkages are operative, include: alg-geom@publications.math.duke.edu (algebraic geometry) astro-ph@babbage.sissa.it (astrophysics) cond-mat@babbage.sissa.it (condensed matter) e-mail@xxx.lanl.gov (e-mail address database) funct-an@babbage.sissa.it (functional analysis) gr-qc@xxx.lanl.gov (gravitation, quantum cosmology) nucl-th@xxx.lanl.gov (nuclear physics, theory) hep-lat@ftp.scri.fsu.edu (computational and lattice physics) hep-ph@xxx.lanl.gov (high energy physics, phenomenological) hep-th@xxx.lanl.gov (high energy physics, formal) A list is also being compiled of other electronic preprint systems (based on other software) in all scientific disciplines. At present the list includes: mp_arc@math.utexas.edu (mathematical physics preprints) cycler@goshawk.lanl.gov (cycle expansions preprints) csp2.csp.uga.edu [128.192.19.229] (simulational physics FTP server) If you know of other such systems, please send e-mail as a "comment" To: nlin-sys@xyz.lanl.gov Subject: comment describing the scope of the system, its internet address, and "help" instructions. Comments on nlin-sys may also be sent to the above address or to any of the category addresses, and will be forwarded if appropriate to the scientific advisors. A copy of this announcement is obtainable by sending e-mail To: nlin-sys@xyz.lanl.gov Subject: get announce {*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*} {*}+ revised 1993.3.12 +{*} {*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*}+{*} ------------------------------ From: Joerg Heitkoetter Date: Sat, 20 Mar 93 16:58:36 +0100 Subject: Announcing the hodge-podge machine... Enclosed the README file of HODGE-C, a cellular-automata simulator. HODGE-C: -------- HODGE-C is a (`mostly ANSI') C language implemenatation of Gerhard & Schuster's hodge-podge machine. It implements a class of cellular automata, that resemble very closely autocatalytic chemical reactions, like for example, the Belousov-Zhabotinskii (BZ) reaction. It's main purpose is given in the following ex- cerpt from the manual page: ANIMATION hodge's main purpose is to generate the previously mentioned snapshot files for `making movies'. All you need is, eg. E. I. du Pont de Nemours & Company's ImageMagick PD software (available via anonymous ftp from export.lcs.mit.edu, as file `contrib/ImageMagick.tar.Z.') using the package's ani- mate(1) command you'll soon see BZ waves move across the grid. An example for this is given in the hodge/movie folder of the HODGE-C distribution. A different approach is to use IDL from Research Systems, Inc. Boulder, CO, USA. An example for this purpose is given in the hodge/idl folder of the HODGE-C distribution. An animation companion to pixmon(1) called playmate(1) is also in the making, and will be distributed in near future (refer to the AVAILABILITY section below). Its application is rather simple: all there is to do is to include a ``tee- log'' in the `--monitor-process' eg.: --monitor-process 'tee >pmdata.pix | pixmon ...' to collect pixmon(1) data, that is later turned into animated pixmaps by playmate(1). A final, powerful and low cost solution for animation is to install the KHOROS visualization package on your system. It comes for free from the University of New Mexico, and is better than most commercial products. KHOROS also features an easy to use graphical programming interface called CAN- TATA and is very easy to learn for beginners in scientific visualization. KHOROS features a bunch of converters to it's internal VIFF(5) format, that is easily generated from hodge's dump (eg. asc2viff(1)) and animation data files (eg. pbm2viff(1), rast2viff, raw2viff(1), etc.) [..] For an introductory article on hodge-podge see: A.K.Dewdney's Computer Recreations column, in Scientific American, August 1988. HODGE-C is distributed free of charge and the other terms of the GPL, ie. the GNU General Public License. HODGE-C v0.98j is available via ftp from lumpi.informatik.uni- dortmund.de (129.217.36.140). Log on with user name "ftp" and give your full e-mail address as password. The file hodge-c- 0.98j.tar.Z in directory /pub/CA/src contains the complete software and documentation. HODGE-C v0.98j was posted to comp.sources.misc, and thus is available from every newsarchive server close to you. (In Ger- many ftp to ftp.germany.eu.net, and see directory /pub/newsarchive/comp.sources.misc) For more information contact: Joerg Heitkoetter (joke@ls11.informatik.uni-dortmunde.de) c/o Systems Analysis Group, LSXI Department of Computer Science //// University of Dortmund UNI DO// P.O. Box 500 500 ___ //// 4600 Dortmund 50 \*\\/// Germany \\\\/ ------------------------------ Date: Mon, 22 Mar 93 19:23:10 GMT From: Toby Tyrrell Subject: Animal Behaviour Software Available PhD dissertation and simulation software available [instructions at end of message] DISSERTATION. Imagine a zebra in the African savannah. At each moment in time this zebra has to weigh up alternative courses of action before deciding which one will be most beneficial. For instance, it may want to graze because it is short of food, or it may want to head towards a water hole because it is short of water, or it may want to remain motionless in order to avoid detection by the predator lurking nearby. This is an example of the problem of action selection: how to choose, at each moment in time, the most appropriate out of a repertoire of possible actions. This thesis investigates action selection in a novel way and makes three main contributions. Firstly, a description is given of a simulated environment which is an extensive and painstaking simulation of the problem of action selection for animals. Secondly, this complex simulated environment is used to investigate the adequacy of several theories of action selection (from both ethology and artificial intelligence) such as the drive model, Lorenz's psycho-hydraulic model and Maes' spreading activation network - deficiencies in all of these mechanisms were discovered and their implications are outlined. Thirdly, a new approach to action selection is developed which determines the most appropriate action in a principled way, and which does not suffer from the inherent shortcomings found in other methods. A review and bibliography of existing work on action selection is included. To obtain the thesis: 1 - "ftp ftp.ed.ac.uk" (129.215.146.5) 2 - user name = "anonymous", password = your user name 3 - "binary" 4 - "cd pub/lrtt" 5 - "get as.1.ps.Z", "get as.2.ps.Z" .... "get as.7.ps.Z" 6 - "quit" and then "uncompress as.1.ps.Z". The postscript file can then be previewed (e.g. "ghostview as.1.ps") or printed (e.g. "lpr -Plw as.1.ps"). Similarly for the other 6 parts. SIMULATION SOFTWARE. This was written in Suntools rather than Xtools. It can be run only from SunView or OpenWindows. The action selection problem modelled by the simulated environment comprises 15 different `sub-problems' (getting food, reproducing, not getting lost, being vigilant for predators, etc), many internal and external stimuli, and 35 different low-level actions to select between. To obtain the software: 1 - "ftp ftp.ed.ac.uk" (129.215.146.5) 2 - user name = "anonymous", password = your user name 3 - "binary" 4 - "cd pub/lrtt" 5 - "get se.tar.Z" 6 - "quit" and then to install it "uncompress se.tar.Z ; tar xvf se.tar" and follow the instructions in as/README. ------------------------------ Date: Tue, 23 Mar 93 14:39:00 -0500 From: hhchou@cs.UMD.EDU (Hui-Hsien Chou) Subject: Announcement of Availability of CA software Simple Systems Exhibiting Self-Directed Replication: Transition Functions, Software, and Documentation March, 1993 We have recently developed and studied cellular automata models of self-replicating systems [Science, 259, 1993, pp. 1282-1288]. Files containing a version of the various transition functions and software that we used in this study, as well as documentation, are now available via ftp. The cellular automata software is actually fairly general and could also serve as an application-independent simulator. For further details please refer to the paper cited above. Here is a brief description of how to get this online information: 1. If your system is not on Internet we cannot help. You might ask your friends on Internet to get this material for you. 2. If you are on Internet, type ftp ftp.cs.umd.edu if this does not work, type ftp 128.8.128.8 You will see the following message Connected to 128.8.128.8. 220 mimsy.cs.umd.edu FTP server (Version 4.135 Thu Apr 26... Name (128.8.128.8:hhchou): Now type in the login name "anonymous", then type in your email id as the password. You will see the following prompt Name (128.8.128.8:hhchou): anonymous 331 Guest login ok, send ident as password. Password: 230 Guest login ok, access restrictions apply. ftp> 3. Type the following command to change directory to /pub cd pub Next type the following command to signify binary file transfer bi You will see the following message ftp> cd pub 250 CWD command successful. ftp> bi 200 Type set to I. ftp> 4. Next, type the following command to get the files. get dtr.tar.Z You will see something like this ftp> get dtr.tar.Z 200 PORT command successful. 150 Opening data connection for dtr.tar.Z (128.8.128.151,... 226 Transfer complete. local: dtr.tar.Z remote: dtr.tar.Z 1598529 bytes received in 49 seconds (32 Kbytes/s) ftp> Type the following command to quit ftp quit 5. After you have transferred the files, move them somewhere in your system where you want to store them. Finally, type zcat dtr.tar.Z | tar xvf - This will retrieve everything in this package, and give you a directory called "dtr". Follow the instructions in the README file there. We hope you find this package useful to you. Note that although care has been taken while preparing this package, we know bugs are there somewhere. We do not have the resources to provide maintenance and support for this software. Suggestions or bug reports are welcome. If time permits we will try to fix any bug found and provide an improved version in the future. We would appreciate it if you acknowledge the use of this software in any future research reports that you might base upon it and send us a copy of such reports. Please send your correspondence to: hhchou@eng.umd.edu. Hui-Hsien Chou Dept. of Computer Science A.V. Williams Bldg. University of Maryland at College Park College Park, MD 20742 USA ------------------------------ From: Inman Harvey Date: Wed, 24 Mar 93 10:17:28 GMT Subject: Evolutionary Robotics - Tech. Reports Evolutionary Robotics at Sussex -- Technical Reports =============================== The following six technical reports describe our recent work in using genetic algorithms to develop neural-network controllers for a simulated simple visually-guided robot. Currently only hard-copies are available. To request copies, mail one of: inmanh@cogs.susx.ac.uk or davec@cogs.susx.ac.uk or philh@cogs.susx.ac.uk giving a surface mail address and the CSRP numbers of the reports you want. or write to us at: School of Cognitive and Computing Sciences University of Sussex Brighton BN1 9QH England, UK. ------------ABSTRACTS-------------------- Genetic convergence in a species of evolved robot control architectures I. Harvey, P. Husbands, D. Cliff Cognitive Science Research Paper CSRP267 February 1993 We analyse how the project of evolving 'neural' network controller for autonomous visually guided robots is significantly different from the usual function optimisation problems standard genetic algorithms are asked to tackle. The need to have open ended increase in complexity of the controllers, to allow for an indefinite number of new tasks to be incrementally added to the robot's capabilities in the long term, means that genotypes of arbitrary length need to be allowed. This results in populations being genetically converged as new tasks are added, and needs a change to usual genetic algorithm practices. Results of successful runs are shown, and the population is analysed in terms of genetic convergence and movement in time across sequence space. Analysing recurrent dynamical networks evolved for robot control P. Husbands, I. Harvey, D. Cliff Cognitive Science Research Paper CSRP265 January 1993 This paper shows how a mixture of qualitative and quantitative analysis can be used to understand a particular brand of arbitrarily recurrent continuous dynamical neural network used to generate robust behaviours in autonomous mobile robots. These networks have been evolved in an open-ended way using an extended form of genetic algorithm. After briefly covering the background to our research, properties of special frequently occurring subnetworks are analysed mathematically. Networks evolved to control simple robots with low resolution sensing are then analysed, using a combination of knowledge of these mathematical properties and careful interpretation of time plots of sensor, neuron and motor activities. Analysis of evolved sensory-motor controllers D. Cliff, P. Husbands, I. Harvey Cognitive Science Research Paper CSRP264 December 1992 We present results from the concurrent evolution of visual sensing morphologies and sensory-motor controller-networks for visually guided robots. In this paper we analyse two (of many) networks which result from using incremental evolution with variable-length genotypes. The two networks come from separate populations, evolved using a common fitness function. The observable behaviours of the two robots are very similar, and close to the optimal behaviour. However, the underlying sensing morphologies and sensory-motor controllers are strikingly different. This is a case of convergent evolution at the behavioural level, coupled with divergent evolution at the morphological level. The action of the evolved networks is described. We discuss the process of analysing evolved artificial networks, a process which bears many similarities to analysing biological nervous systems in the field of neuroethology. Incremental evolution of neural network architectures for adaptive behaviour D. Cliff, I. Harvey, P. Husbands Cognitive Science Research Paper CSRP256 December 1992 This paper describes aspects of our ongoing work in evolving recurrent dynamical artificial neural networks which act as sensory-motor controllers, generating adaptive behaviour in artificial agents. We start with a discussion of the rationale for our approach. Our approach involves the use of recurrent networks of artificial neurons with rich dynamics, resilience to noise (both internal and external); and separate excitation and inhibition channels. The networks allow artificial agents (simulated or robotic) to exhibit adaptive behaviour. The complexity of designing networks built from such units leads us to use our own extended form of genetic algorithm, which allows for incremental automatic evolution of controller-networks. Finally, we review some of our recent results, applying our methods to work with simple visually-guided robots. The genetic algorithm generates useful network architectures from an initial set of randomly-connected networks. During evolution, uniform noise was added to the activation of each neuron. After evolution, we studied two evolved networks, to see how their performance varied when the noise range was altered. Significantly, we discovered that when the noise was eliminated, the performance of the networks degraded: the networks use noise to operate efficiently. Evolving visually guided robots D. Cliff, P. Husbands, I. Harvey Cognitive Science Research Paper CSRP220 July 1992 We have developed a methodology grounded in two beliefs: that autonomous agents need visual processing capabilities, and that the approach of hand-designing control architectures for autonomous agents is likely to be superseded by methods involving the artificial evolution of comparable architectures. In this paper we present results which demonstrate that neural-network control architectures can be evolved for an accurate simulation model of a visually guided robot. The simulation system involves detailed models of the physics of a real robot built at Sussex; and the simulated vision involves ray-tracing computer graphics, using models of optical systems which could readily be constructed from discrete components. The control-network architecture is entirely under genetic control, as are parameters governing the optical system. Significantly, we demonstrate that robust visually-guided control systems evolve from evaluation functions which do not explicitly involve monitoring visual input. The latter part of the paper discusses work now under development, which allows us to engage in long-term fundamental experiments aimed at thoroughly exploring the possibilities of concurrently evolving control networks and visual sensors for navigational tasks. This involves the construction of specialised visual-robotic equipment which eliminates the need for simulated sensing. Issues in evolutionary robotics I. Harvey, P. Husbands, D. Cliff Cognitive Science Research Paper CSRP219 July 1992 In this paper we propose and justify a methodology for the development of the control systems, or `cognitive architectures', of autonomous mobile robots. We argue that the design by hand of such control systems becomes prohibitively difficult as complexity increases. We discuss an alternative approach, involving artificial evolution, where the basic building blocks for cognitive architectures are adaptive noise-tolerant dynamical neural networks, rather than programs. These networks may be recurrent, and should operate in real time. Evolution should be incremental, using an extended and modified version of genetic algorithms. We finally propose that, sooner rather than later, visual processing will be required in order for robots to engage in non-trivial navigation behaviours. Time constraints suggest that initial architecture evaluations should be largely done in simulation. The pitfalls of simulations compared with reality are discussed, together with the importance of incorporating noise. To support our claims and proposals, we present results from some preliminary experiments where robots which roam office-like environments are evolved. ------------------------------ End of ALife Digest *******************

---

E-Mail Fredric L. Rice / The Skeptic Tank