From owner-gann-list Wed Apr 3 02:05:03 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id BAA11686 for gann-list-outgoing; Wed, 3 Apr 1996 01:48:32 -0600 (CST) From: GANN List Admin Message-Id: <199604030748.BAA19973@ren.cs.iastate.edu> Subject: GANN: Reminder. To: gann-list@cs.iastate.edu Date: Wed, 3 Apr 1996 01:48:23 -0600 (CST) X-Mailer: ELM [version 2.4 PL25] Content-Type: text Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: GANN List Admin Dear Colleagues, Please read the following notes on the operation of the GANN list and save them for future reference, if you haven't already done so. Thank you for your cooperation in this matter. - GANN Admin - _______________________________________________________________________ To send an article or note to the subcribers of this list: send the message to -- gann-list@cs.iastate.edu For administrative queries/suggestions/comments send e-mail to -- gann-list-owner@cs.iastate.edu To unsubscribe: send e-mail to majordomo@cs.iastate.edu with *body* (not subject) of: unsubscribe gann-list [optional e-mail addr to be unsubscribed] To subscribe: send e-mail to majordomo@cs.iastate.edu with *body* (not subject) of: subscribe gann-list [optional e-mail addr to be subscribed] GANN Home page: There is a homepage for GANN, which can be accessed through a suitable WWW browser like Mosaic, Netscape etc. at the URL: http://www.cs.iastate.edu/~gannadm/homepage.html GANN DIGESTS ARCHIVE: "digests" of messages sent out on this list is available through majordomo (old digests are available through ftp). To retrieve file names via majordomo, send mail to majordomo@cs.iastate.edu with the command "index gann-list". To retrieve an archive, send mail with the command "get gann-list " where is the name of the archive you wish to retrieve from the list returned by the "index" command. "Old" digests through ftp: unix prompt% ftp ftp.cs.iastate.edu login: anonymous password: unix prompt% cd pub/gann/gann-digest-archive unix prompt% get Thank you for your interest in the GANN Mailing List. _______________________________________________________________________ Dr. Vasant Honavar Karthik Balakrishnan Assistant Professor Doctoral Student Artificial Intelligence Research Group Department of Computer Science Iowa State University Ames, IA-50011, USA. honavar@cs.iastate.edu balakris@cs.iastate.edu Dr. Mike Rudnick Adjunct Assistant Professor Department of Computer Science and Engineering Oregon Graduate Institute rudnick@cse.ogi.edu _______________________________________________________________________ From owner-gann-list Wed Apr 3 06:42:25 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id GAA26260 for gann-list-outgoing; Wed, 3 Apr 1996 06:38:50 -0600 (CST) Date: Wed, 3 Apr 1996 14:39:28 +0200 (MET DST) Message-Id: <199604031239.OAA13092@alto.unice.fr> From: Olivier MICHEL To: GANN-List Subject: GANN: Khepera Contest Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: Olivier MICHEL Here is the call for contest of Khepera Contest which will be held in Nimes, France, during Fall 1997. Please note that it takes often a lot of time for the competitors to prepare such a contest (organization of student projects, research developments), that's why this is an early announcement to let you enough time to be good competitors. So don't wait too much if you want to have a chance to win a real Khepera... Olivier MICHEL. om@alto.unice.fr http://wwwi3s.unice.fr/~om/ PS: I apologize if you already received this message through various other lists / newsgroups. --8<--------------------------------------------------------------------------- Call for contest: --------------------------------------------------------------------- # Khepera Contest # --------------------------------------------------------------------- Write your own adaptive robot controller on the simulator, and win a Khepera robot INTRODUCTION This contest intends to confront various controller systems driving the mobile robot Khepera and exhibiting adaptive behavior. A Khepera simulator sofware is provided to the competitors. The live confrontation will occur on both real robot and simulator during Evolution Artificielle conference in Nimes, 1997. The first prize will be a Khepera robot (or an extension module for Khepera, if the winner already owns a Khepera). CONTEST GOAL To build up the most interesting robot controller that exhibit adaptive behavior, using learning or evolutionary methods (or both). CONSTRAINTS All possible forms of controller are accepted including: Artificial neural networks. Classifiers systems. Subsumption architectures. Fuzzy logic systems. Blackboard based systems. Anything else you can imagine. Evolutionary and learning methods are: Genetic algorithms. Evolutionary Programming. Reinforcement learning. Bucked brigad algorithm. Any other evolutionary or learning technique you just designed. EVALUATION During Evolution Artificielle conference, the controllers will be run on the simulator and on the real robot. The competitors should explain the techniques they used (posters are welcome). The jury will consider several aspects of the system including: Originality. Overall Efficiency. Relevance to the evolutionary methods. Relevance to the learning methods. CATEGORIES We will make a distinction between Khepera owners and other competitors, who have only the simulator to develop their system. Consequently, the jury will be a bit more indulgent with the latter. PRIZES First prize: Choice between a Khepera robot or an extension module for Khepera robot. Second prize: A Khepera bag containing Valentino Braitenberg's book: Vehicles. Third prize: An official Khepera T-shirt. IMPORTANT DATES June 1997: Contest registration deadline. October 1997: Live Contest at Evolution Artificielle Conference in Nimes, jury deliberation and prizegiving. Don't wait too much, register and start now ! Student research projects and professional researcher works are welcome. REGISTRATION and SOFTWARE DOWNLOADING Registration and Software downloading are available on the World Wide Web at: http://wwwi3s.unice.fr/~om/khep-contest.html KHEPERA INFORMATION Khepera is a mini mobile robot developped at EPFL by Edo Franzi, Andre Guignard and Francesco Mondada (K-Team). More information is available at: http://lamiwww.epfl.ch/Khepera/ ORGANIZATION COMMITEE o Olivier Michel, Laboratoire i3S, CNRS, University of Nice, F o Francesco Mondada, K-Team SA, Preverenges, CH o Marc Schoenauer, Ecole Polytechnique Paris, F From owner-gann-list Tue Apr 9 10:33:28 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id KAA02328 for gann-list-outgoing; Tue, 9 Apr 1996 10:04:01 -0500 (CDT) From: DPATINO@inaut.edu.ar (Daniel Pati¤o) To: gann-list@cs.iastate.edu Date: Tue, 9 Apr 1996 11:14:23 Subject: GANN: Doctoral thesis available Priority: normal X-mailer: Pegasus Mail v3.1 (R1a) Organization: Instituto de Automatica (INAUT) Message-ID: <9604091200.aa15105@unsjr.unsj.edu.ar> Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: DPATINO@inaut.edu.ar (Daniel Pati¤o) The following Doctoral Thesis in Engineering (Control Systems Engineering Program) is available via the Instituto de Automatica (prompt by internet). Dynamic Control of Robot Manipulators Using Neural Networks by Hector Daniel Patino June 27, 1995 Advisors: Profs. Benjamin Kuchen and Ricardo Carelli Abstract: Robotics, as a branch of Automation, has become a very important subject of modern production plants. The present thesis is developed in this field, within the frame of Control System Engineering. It specifically deals with the development and analysis of neural network- based controllers for motion control of robot manipulators. The need to control increasingly complex systems having high- performance and restricted tolerances, such as robots, and considering uncertainties both in its mathematical model and in the interacting environment, has encouraged the use of new control strategies. Artificial Neural Networks (ANNs) provide an alternate approach for solving the cited problem, which can hardly be faced with traditional methodologies. This thesis deals with neural network-based controllers for non- adaptive and adaptive motion dynamic control of robot manipulators. Much attention has been put on ANN-based controllers in recent years. This type of controller exploits the possibilities of ANNs for learning nonlinear functions through experimental data, as well as for solving certain types of problems where massive parallel computation is required. However, most different schemes published so far, have been proposed without any rigorous stability analysis nor any consideration on the influence of ANN learning errors on motion control errors. The dynamic behaviour of a rigid manipulator can be modelled by a system of highly coupled and nonlinear differential equations. The nonlinear effects become more significant in high performance robots which operate with direct drives or reduced gains in transmission gears, and the mass of the manipulated load is considerably larger than that of the robot structure or the operating speed is large. In this thesis, neural networks- based structures for modelling the dynamic behaviour of robot manipulators are considered and proposed. These structures take advantage of the ANNs capacity for learning the nonlinear dynamic model of the robot through experimental data. New structures are presented for non-adaptive and adaptive motion control of high performance robot manipulators based on ANNs: Neural network-based feedforward and feedback controllers. The adaptive controllers include a set of off-line trained ANNs (fixed neural networks) and an update law for adjusting the robot's inverse dynamics. In a second approach, the ANN structure is defined in such a way that the synapsis of the last layer represents specifically the dynamic parameters and payload uncertainties of the parameterization system model. These structures allows to adapt the controller to dynamic uncertainties such as link inertias or payloads, thus minimizing the amount of on-line computation. Since the number of parameters to be adjusted is smaller, the adaptation to changes in robot parameters is faster than by using the learning capabilities of the complete ANN to adapt, allowing the on-line adaptation. It is also presented a procedure for training each ANN of the set. In all phases, ANNs are trained using the backpropagation learning rule, with its learning rate varying in response to a fuzzy set system. This process accelerates the learning time and restricts the chances of falling into a local minimum. Considering the ANN learning error, strong practical stability conditions are given for the proposed controllers. The analysis allows a control error evaluation as a function of the ANN learning errors. The evaluation functions set the explicitly functional relationship between the control errors and ANN learning error and the design parameters. These expressions are very useful also from a practical point of view. In addition, robust controllers to the ANN learning errors are proposed, using a sign or saturation switching function in the control laws, which lead to global asymptotic stability and zero convergence of control errors. In order to verify the stability properties and performance of the proposed control schemes, simulation studies were carried out using the PUMA-560 robot model, which ease the transition to an experimentation with industrial robots. Laboratory experimental results were obtained for neural network-based feedforward non- adaptive and adaptive controllers. The experimental results demonstrate the practical feasibility and satisfactory performance of neural network-based identification and the related control system. Lab experiences were performed on a one-degree-of-freedom robot with direct driver (motorized pendulum), and the results confirmed those obtained from theoretical analysis and simulations. Summary on main contributions of the present work: an alternative to overcome some drawbacks found in advanced control of high- performance robots. The neural network-based feedforward and feedback controllers here proposed, for non-adaptive and adaptive motion dynamic control of high performance robot manipulators, take advantage of the capacity of ANNs for learning the robot model through experimental data, as well as its operative mode, that is, massive parallel computation, becoming an adequate tool for real-time control. These advantages offer an alternate solution for developing advanced adaptive controllers of high-performance robots. In case of adaptive controllers and in presence of uncertainties in payload and other dynamic parameters, the feedforward and feedback signals are adapted according to a parameter updating law. The adaptive process is much faster than that of using the learning capabilities of the entire ANN array, thus decreasing the amount of on-line computation. The neural network-based motion adaptive controllers for robot manipulators use a bank of fixed neural networks, connected in feedforward or feedback configuration plus a lineal PD controller. Each of the ANNs is off-line trained with adequate payload conditions, and they develop the dynamic model following a sort of look-up table approach, doing interpolations between the models corresponding to particular load conditions. This way for modelling the inverse dynamics of the robot through neural networks enable to capture unmodeled dynamics by input-output experimental data, and taking into account the robot's full dynamics as well as any probable disturbance or unwanted dynamic effect such as non-linear friction, little elasticity in joints and links, backlash, etc. Furthermore, the main advantage of the ANNs bank is that it avoids using the parametric model based dynamics. Indeed, there is no need to estimate the parameters of the system, as is the case of the traditional direct and indirect adaptive control. Here, it is only necessary to increment the number of ANNs of the bank, thus preventing the need of a priori knowledge of the robot dynamic model. Due to the neural networks learning error, practical asymptotic stability conditions are given for the proposed controllers. The control objective is accomplished by reaching a sufficiently small tracking motion error and keeping all other signals bounded. The analysis also allows an evaluation of the control error bound as an explicit function of ANN learning errors and design parameters. Although the main disadvantages are related to the number of nodes and number of bank of ANNs, the learning error of ANN can be evaluated, thus enabling to estimate the tracking error bounds. In addition, feedforward or feedback adaptive controllers, which are robust to ANN learning errors, are proposed, using a sign or saturation switching function in the control law, which leads to global asymptotic stability and zero convergence of tracking control errors. Simulations and experimental results, after applying the controllers to the motion control of two joints of the PUMA-560 robot, verified the stability properties obtained in theory, and demonstrated the practical feasibility of these controllers, and their satisfactory tracking performance. As a conclusion, the experience and results attained through the present work suggest the possibility for expanding this process to neural network-based modelling, identification and control of other systems and processes. Current address: Instituto de Automatica, Facultad de Ingenieria, Universidad Nacional de San Juan, Av. San Martin 1109 Oeste, (5400) San Juan, ARGENTINA. E-mail: dpatino@inaut.edu.ar. Fax: +54 64 213672 (Searching a postdoctoral fellow) From owner-gann-list Fri Apr 19 17:49:11 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id RAA02600 for gann-list-outgoing; Fri, 19 Apr 1996 17:31:28 -0500 (CDT) Date: Tue, 16 Apr 1996 12:05:48 GMT From: stefano@kant.irmkant.rm.cnr.it (Stefano Nolfi) Message-Id: <9604161205.AA19378@kant.irmkant.rm.cnr.it> To: alife@cognet.ucla.edu, connectionists@cs.cmu.edu, gann@cs.iastate.edu Subject: GANN: Paper available on adaptive classification with autonomous robots Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: stefano@kant.irmkant.rm.cnr.it (Stefano Nolfi) Paper available via WWW / FTP: Keywords: Active Perception, Adaptive Behaviors, Evolutionary Robotics, Neural Networks, Genetic Algorithms. ------------------------------------------------------------------------------ ADAPTATION AS A MORE POWERFUL TOOL THAN DECOMPOSITION AND INTEGRATION Stefano Nolfi Institute of Psychology, C.N.R., Rome. Recently a new way of building control systems, known as behavior based robotics, has been proposed to overcome the difficulties of the traditional AI approach to robotics. Most of the work done in behavior-based robotics involves a decomposition process (in which the behavior required is broken down into simpler sub-components) and an integration process (in which the modules designed to produce the sub-behaviors are put together). In this paper we claim that decomposition and integration should be the result of an adaptation process and not of the decision of an experimenter.To support this hypothesis we show how in the case of a simple task in which a real autonomous robot is supposed to classify objects of different shapes, by letting the entire behavior emerge through an evolutionary technique,a more simple and robust solution can be obtained than by trying to design a set of modules and to integrate them. http://kant.irmkant.rm.cnr.it/public.html or ftp-server: kant.irmkant.rm.cnr.it (150.146.7.5) ftp-file : /pub/econets/nolfi.recog.ps.Z for the homepage of our research group with most of our publications available online and pointers to ALIFE resources see: http://kant.irmkant.rm.cnr.it/gral.html ---------------------------------------------------------------------------- Stefano Nolfi Institute of Psychology, C.N.R. Viale Marx, 15 - 00137 - Rome - Italy voice: 0039-6-86090231 fax: 0039-6-824737 e-mail: stefano@kant.irmkant.rm.cnr.it www: http://kant.irmkant.rm.cnr.it/nolfi.html From owner-gann-list Sat Apr 20 09:10:36 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id JAA06111 for gann-list-outgoing; Sat, 20 Apr 1996 09:03:33 -0500 (CDT) From: R Roy Organization: University of Plymouth To: r.roy@ieee.org, engineering-all@mailbase.ac.uk, engineering-design@mailbase.ac.uk, evolutionary-computing@mailbase.ac.uk, postgrad@mailbase.ac.uk, Connectionists@cs.cmu.edu, hybrid-list@cs.ua.edu, ga-list@aic.nrl.navy.mil, gann-list@cs.iastate.edu, ml@ics.uci.edu, INDUCTIVE@hermes.csd.unb.ca, nafips-l@sphinx.Gsu.EDU, neuron-request@CATTELL.PSYCH.UPENN.EDU, fuzzy-mail@dbai.tuwien.ac.at, design-research@mailbase.ac.uk, conferences@iao.fhg.de, cti-comp-ai@mailbase.ac.uk, cti-engineering@mailbase.ac.uk, defence-all@mailbase.ac.uk, defence-technology@mailbase.ac.uk, design-research@mailbase.ac.uk, tanmay@naval.iitkgp.ernet.in, R.Roy@plymouth.ac.uk, R.Bapi@plymouth.ac.uk Date: Sat, 20 Apr 1996 15:02:55 BST Subject: GANN: SHOT'97 Priority: normal X-mailer: Pegasus Mail v3.30 Message-ID: MIME-Version: 1.0 Content-type: text/plain; charset="US-ASCII" Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: R Roy FIRST ANNOUNCEMENT CALL FOR PAPERS CONFERENCE ON SHIP AND OCEAN TECHNOLOGY SHOT '97 March 5-7, 1997 Department of Ocean Engineering and Naval Architecture Indian Institute of Technology Kharagpur 721 302 INDIA CONFERENCE The department of Ocean Engineering and Naval Architecture at the Indian Institute of Technology, Kharagpur established in 1951 is the first of its kind in India in imparting education, research and training in naval architecture in the country. Its scope has subsequently been increased to include ocean engineering. The facilities in the department have increased over the years and various experimental and theoretical research projects have been and are being conducted on various aspects of naval architecture and ocean engineering. The main aim of the conference is to discuss the present status, latest development and future trends in the analysis, design, research, manufacture and maintenance of various ocean engineering structures and vehicles and related technological aspects. It is intended to organise an exibition of industrial products, computer software related to Ship and Ocean industries and also a Poster Session. CONFERENCE TOPICS Technical Papers are invited on the following broad topics. * Design and production of marine vehicles and systems * Submerged and deep sea vehicles * High performance and novel marine crafts * Marine/Ocean hydrodynamics * Marine environment * Marine/Offshore structures * Marine materials and fabrications * Design and production of ocean systems * CAD/CAM and information system in marine/ocean industry * Managment and maintenance of coastal/ocean structures * Navigation and ship behaviour * Modelling techniques * Testing, instrumentation and measurement techniques CALL FOR PAPERS Prospective authors are invited to submit three copies of 300 word abstract to chairman, Organising Committee. Abstracts must bear the full title of the paper and mailing address, telephone no, Fax and e- mail address of the corresponding author. The authors should indicate the theme in which his paper is to be included. IMPORTANT DEADLINES _ Submission of abstracts: 30 June 1996 _ Notification of abstract acceptance: 30 September 1996 _ Submission of final papers: 30 November 1996 _ Notification of acceptance: 31 December 1996 _ Date of final registration: 31 January 1997 PUBLICATION OF PROCEEDINGS Papers presented in the conference will be published in the proceedings. REGISTRATION FEES Participant From India Outside India Authors Rs. 1000.00 US$ 300.00 Other Delegates Rs. 1000.00 US$ 300.00 Accompanying Person Rs. 500.00 US$ 150.00 MODE OF PAYMENT OF FEES Payment of registration fees may be made by demand draft or by international money order or by bank transfer. The drafts is to be drawn in favour of Conference on Ship and Ocean Technology SHOT'97 . Please indicate "Conference on Ship and Ocean Technology" and your name in all payments. Please mail the draft/money order to the Chairman, SHOT'97, Department of Ocean Engineering and Naval Architecture, Indian Institute of Technology, Kharagpur 721 302, India. ACCOMODATION Kharagpur is 117 kilometers away from Calcutta. Kharagpur is connected by rail. The month of March is very pleasant at Kharagpur. Temperature is likely to range from 27-330C. All the delegates of the conference will be provided accomodation in the guest houses at the Institute. The tariff /day/person for the Institute Guest House is Rs. 240 for A.C. and is Rs. 90 for non- A.C. and available on first come first serve basis. The charges for CEC guest is Rs. 125 for A.C. and Rs.40 for non A.C. (double occupancy). ACCOMPANYING PERSONS Persons accompanying delegates who are not attending technical sessions are also welcome. On prior intimation accommodation will be arranged for accompanying persons. It is also planned to arrange suitable engagements for such persons depending on response. Suitable additional charges will be levied for this purpose. PRE REGISTRATION FORM The Conference on Ship and Ocean Technology Department of Ocean Engineering and Naval Architecture Indian Institute of Technology Kharagpur 721 302 West Bengal, INDIA Name: Designation: Conference Attendence: Wish to attend conference Present a paper Wish to bring accompanying persons Mailing Address: Fax: e-mail: Telephone: Telex: Signature ====================================================================== The Pre registration form may be sent to Organising Chairman SHOT 97 Department of Ocean Engineering and Naval Architecture, Kharagpur 721 302, INDIA Telex: 06401-201 ITKG IN, Telephone; +91 (3222) 2221 - 2224 (4 lines) Extn. 4460 FAX: +91 (3222) 2303 e-mail: misra@naval.iitkgp.ernet.in ====================================================================== ********************************************************************** Rajkumar Roy Plymouth Engineering Design Centre Charles Cross Centre University of Plymouth Plymouth PL4 8AA UK Email : r.roy@ieee.org, rroy@plymouth.ac.uk Tel. : 01752 233508 ********************************************************************** From owner-gann-list Fri Apr 26 10:14:59 1996 Received: (from mdomo@localhost) by cs.iastate.edu (8.7.4/8.7.1) id JAA18850 for gann-list-outgoing; Fri, 26 Apr 1996 09:42:54 -0500 (CDT) Date: Fri, 26 Apr 96 15:30:25 BST From: Noel Sharkey Message-Id: <9604261430.AA09596@dcs.shef.ac.uk> To: connect-bb@ed.eusip, incon@dcs.shef.ac.uk, reinforce@cs.uwa.edu.au, gann-list@cs.iastate.edu, neuron-request@CATTELL.psych.upenn.edu, alife@cognet.ucla.edu, cogpsy@neuro.psy.soton.ac.uk, connectionists@CS.CMU.EDU, epsynet@uhupvm1.bitnet, elsnet-list@cogsci.ed.ac.uk, irl-net@irlearn.bitnet, arpanet-bboards@edu.mit.lcs.mc, hybrid-list@cs.ua.edu, colt@cs.uiuc.edu, ilpnet@ijs.si, cphc-jobs@ukc.ac.uk Subject: GANN: 3 Research Studentships available Cc: A.Sharkey@dcs.shef.ac.uk Sender: owner-gann-list@cs.iastate.edu Precedence: bulk Reply-To: Noel Sharkey PLEASE PASS ON TO ANY FINAL YEAR UNDERGRADUATES OR MASTERS WHO ARE SEEKING FUNDED PHD PLACES. Sorry if you receive this more than once. 3 RESEARCH STUDENTSHIPS IN NEURAL COMPUTING AND ROBOTICS Department of Computer Science University of Sheffield Three funded PhD studentships are available, two from from 1st August, and one from end of September, 1996. The first of these is restricted to British Students only and the other 2 are for students from countries within the European Community. 1. Neural Computing. Projects on any topic within the field of neural computing will be considered. Two areas of particular interest are (a) Improving the reliability of neural computing applications through the use of ensembles of nets and (b) Cognitive modelling, transfer and interference. 2. Autonomous Mobile Robotics. The project would provide an ideal opportunity for a creative student to work on the development of ``intelligent'' behaviour on a mobile robot. Neural computing techniques have already been applied in our lab to develop a number of low level behaviours on a Nomad200. The student would be expected to develop higher level behavioural control. There are a number of different approaches that could be taken. For example: using representations developed at the lower levels to induce higher level behaviours or using a human and animal developmental paradigm. However, nothing is set in stone for this project and a good proposal will go a long way. 3. Pharmaceutical Robotics Aim: The development of a neural computing system for coordinating robot arms in the task of mixing dangerous drugs. This project is in collaboration with the Pharmacy Unit at the Northern General Hospital. Their problem is that they currently employ more than twenty highly-qualified specialist staff to spend a large part of their day involved in the rather tedious task of mixing drugs. Since many of the drugs are very dangerous to humans (such as anti-cancer drugs), much of the work has to take place inside a sealed glass case that is accessed by attached gloves (a glove box). The solution is to put robot arms into the cases and let them do most of the work. It should be noted that this is a research project and offers a number of interesting robotics problems. The student would not be expected to develop a commercial system. Further information about neural computing within the Artificial Intelligence and Neural Networks (AINN) research group can be viewed on WWW: http://www.dcs.shef.ac.uk/research/groups/nn (This will not be ready to view until Wednesday, 1st, May). Application forms may be obtained from our PhD admission secretary Jill Martin jill@dcs.shef.ac.uk. Or write to Ms J. Martin, Department of Computer Science, 211 Portabello St., Sheffield, S1 4DP, S. Yorks, UK. Forms should be accompanied by a short proposal (less than a page) about what the applicant would like to work on, but this does not commit the applicant.