Format of this file:

____________________________________________________________

Title of paper

Authors

Affiliation

Abstract

Keywords

Place of publication

Filename of PostScript file

____________________________________________________________

WALKING ROBOT ADAPTATION TO GROUND PROPERTIES USING FUZZY CONTROL

Glauco A. de Paula Caurin, Sjur J. Vestli

Institute of Robotics, Swiss Federal Institute of Technology
8092 Zurich, Switzerland

Abstract: The RoboTRAC of the Institute of Robotics, Swiss Federal
Institute of Technology, is an intelligent mechanical digger. This
type of machines are already in use in agriculture and forestry.
The overall project goal is to integrate intelligence and appropriate
sensing capabilities into an automatic controller so that the operation
of the machine is simplified. An automatic controller adapts the walking
patterns based on tactile ground information. A fuzzy logic based
algorithm achieves this adaptation. In simulations this fuzzy controller
(FC) has demonstrated robustness to varying ground properties.The FC
adapts the standard walking pattern based on the history of previous
steps so that the next step sequence will take place with a minimum
of failures. The performance of this controller is demonstrated for
a variety of situations.

Keywords: Walking Machine, Terrain Interaction, Fuzzy Logic.

Published at : IFAC/IMACS/IEEE/IUTAM Workshop on Motion Control
for Intelligent Automation Perugia, October 27-29, 1992

Filename: robotrac_ifac_92.ps
____________________________________________________________

FUN: Optimization of Fuzzy Rule Based Systems Using Neural Networks

Sandra Sulzberger, Nadine Tschichold-Gurman, Sjur J. Vestli

Institute of Robotics, Swiss Federal Institute of Technology
8092 Zurich, Switzerland

Abstract: This paper describes a method for optimization of fuzzy 
rule based systems using neural networks. A new neural network model 
with special neurons has been developed so that the translation of 
fuzzy rules and membership functions into the network is possible. 
The performance of this network and hence the quality of the original 
rule base is then improved by training the network using a combination 
of neural network learning algorithms. The optimized rules and membership 
functions can be extracted from the net and used in normal fuzzy inference 
tools. This net has been tested on the "WallJumperOver" and the problem of 
local navigation for mobile robots

Keywords: Rule base optimisation, Fuzzy Logic.

Published at : IEEE Conference on Neural Networks, San Francisco,
USA, March - April, 1993.

Filename: fun_ieee_93.ps
____________________________________________________________

Realtime Systems for Mechatronics

D. Diez, G. Schweitzer

Institute of Robotics and Mechatronics Lab, ETH Zurich
CH-8092 Zurich, Switzerland
diez@ifr.ethz.ch

Abstract: The paper describes a programming environment
for the software of mechatronic systems. A mechatronic
system implies that a mechanical process with velocities
and forces, i.e. a dynamical process has to be controlled.
The consequences for the computer system are that the time
constraints of the dynamical process have to be met, and
these time constraints are not only desirable but they
are essential and necessary for the correctness of the
program. The object oriented programming style, which is
used in this paper, pushes these time factors and the physical
process into the foreground. The user can concentrate his
attention to the process itself, the programming becomes
a configuration task. An application to robotics explains
how this can be achieved. The task is to build a controller
for a modular robot system. The controller has to be configurable
allowing a new robot type to be set up without change of the
system software. All what is needed is new data for the
description module of the robot. With this approach new
robot applications and automation steps in areas like industrial
assembly, construction work, medicine, service, and research
can be implemented economically and quickly.

Keywords: Object Oriented, Real Time systems.

Submitted for : Second Conference on Mechatronics and Robotics,
Duisburg/Moers (Germany), September 27-29, 1993.

Filename: duisburg_93.ps
____________________________________________________________
