Skip to main content

Über dieses Buch

The areas of intelligent machines or robotic systems is of enormous technological and economic interest as competition in productivity intensifies. This volume gives the proceedings of the 1990 Advanced Study Institute on Expert Systems and Robotics. It presents research work already accomplished in the analytical theory of intelligent machines, work in progress and of current interest and some specific examples for further research. The papers in the volume range from the most theoretical to some descriptions of very practical working robots. The papers are organized into sections on vision and image analysis, robotic sensory systems, software/hardware and system simulation, robot control, applications, and reports of group meetings.



Vision and Image Analysis

Gestalt-based Approach to Robot Vision

The human visual system is usually able to recognize objects as well as their spatial relations without the support of depth information like stereo vision. For this reason we can easily understand cartoons, photographs and movies. It is the aim of our current research to exploit this aspect of human perception in the context of computer and robot vision. From a monocular TV image we obtain information about the type of an object observed in the scene and its position relative to the camera (viewpoint). This report deals with the theory of human image understanding as far as used in this system and describes the realization of a blocks world vision system based on these principles.

Philipp W. Besslich, Henning Bässmann

Matching: Check Points within the Framework of a Knowledge-based Visual Inspection System

Matching within an inspection system is one of the most crucial processing steps in testing how well measures, shape or completeness of an industrial machined work-piece fits the important properties of its “golden prototype”.Especially, pure image based systems suffer from the fact that pixel-to-pixel comparisons are effected by variations in scale, illumination, positioning accuracy or tolerances of single properties of the specimen. Additionally, unprecisely defined features of the test patterns, incomplete specifications of all details to be tested, restrict automation of visual inspection still to few applications only. Knowledge-based matching using a priori information about specimen (e.g. CAD data) and inspection environment would overcome these disadvantages.This paper presents an overview about matching algorithms in knowledge-based visual inspection. From this survey a list of assessment criteria for matching schemes is derived. Initial ideas for a generic matching concept using graph matching techniques is proposed. Discussion of the new approach is mainly devoted to control of matching by prediction graphs derived from model data.

Matthias F. Carlsohn, Lutz Gruber

Stereoscopic Vision for AGV Guidance

This paper first presents the complete scenario for Autonomously Guided Vehicle (AGV) guidance based on stereoscopic vision. The case of indoor environment is examined in details, and the required operations are further discussed. The concept of logical sensors is followed to address the motion parameter estimation problem as a perception task that can be achieved directly by using stereo — camera data. The mathematics for the operations involved are presented focusing on uncertainty issues associated with the physical and mathematical models used.

B. Dimitriadis, K. Papanikolaou

Hierarchical Methods for Robot Vision

There is a large variety of tasks that require robots to sense, to see, or to hear their environment. Even if such a capability is not absolutely necessary, it enables the robot to react on events from the ‘outside’-world. From relatively simple tasks such as inspection up to navigation in an unknown surrounding, vision faces two major problems: the huge amount of data and the need for quick and reliable answers.Massively parallel data and processing structures are required. Among those, image pyramids have shown great robustness and high performance. Originally intended as an ordered collection of images at multiple resolutions, many computations using an image of n × n picture elements need on the order of O(log n) parallel steps. Besides numerical computations, symbolic processing can also be done to extend the application area to mid- and high-level vision tasks. They give us hope to achieve acceptable performance in vision tasks that have to be processed with millions of bytes in an extremely short time.Curves represent shapes at an intermediate level. Hierarchical symbolic curve representations aim at a stepwise reduction of the data to ‘significant parts’, like corners or curvature extrema, while preserving the major property of a curve: its connectivity. The proposed representation and symbolic computation scheme allows fast object recognition and may find applications such as in multi-scale path representation and robot guidance.

Walter G. Kropatsch

Robotic Sensory Systems

Geometric Modeling Aspects on Vision and Ultrasonic Range Sensing for Robotic Applications

In this paper, geometric modelling aspects related with the use of a pair of vision cameras and ultrasonic sensors mounted on a six degrees-offreedom manipulator are described. The information for both types of sensors is used to generate the depth map of the environment. An approach is described to obtain the transformation which relates a coordinate frame associated with an ultrasonic system and the tool coordinate frame. Geometrical aspects related to the use of the ultrasonic system in tasks such as the evaluation of planar-surface equations and manipulator surface following are described, as well. In another section, a method used to calibrate the vision cameras and its optimization by recalibration is presented. The recalibration procedure is made by exploiting the capability of positioning the stereo set in different positions within the workspace.

Urbano Nunes, Jorge Dias, A. Traça de Almeida

Sensor Planning for the Error Diagnosis of Robot Assembly Tasks

A knowledge based system for the diagnosis of assembly process errors is presented. It is part of the action monitoring system of the autonomous mobile assembly robot KAMRO. The diagnosis is supported by the multi-sensory system of the KAMRO system. The system for error diagnosis has the structure of a blackboard system. The task of the sensor system is focused to the monitoring of a product assembly. A very interesting part of the system is an approach for the automatic sensor planning for the acquisition of lacking sensor data. The first prototype of the system is being implemented with the help of the LISP based generic blackboard development shell GBB on a SUN4 computer.

Jörg Raczkowsky, Kerstin Seucken

Software/Hardware, System Simulation

Computer Aided Analysis of Robot Arm Dynamics and Adaptive Control Schemes

A computer program is developed where computer oriented algorithms for robot arm dynamics are implemented together with suboptimal control algorithms for trajectory tracking. The program package allows in an interactive way the synthesis of the manipulator structure as well as its control system, and the simulation of the dynamic behaviour of the whole system under different scenarios. Two adaptive control schemes based on minimum energy terminal control are proposed and the performances of the proposed algorithms are tested with simulation studies.

A. Akmehmet, A. Kuzucu

A Programming Environment for Sensor-Controlled Robots

This paper gives an overview of the research activities in robotics at the Electronics Laboratory of the State University of Ghent over the last few years. Two themes emerge, namely (i) the use of concurrent high-level programming languages to implement the hard real-time control functions needed for sensor driven and compliant motion; and (ii) the need for interactive graphical simulation and programming environments for sensor-equipped robots. With respect to the first theme, we describe the implementation of compliant motion on a Modula-2 based multiprocessor. Regarding the second theme, we observe that environments for programming a robot have been developed ever since robots have existed. These environments have evolved from simple teach-by-doing procedures towards implicit task-level programming and simulation systems. Although some of these complex systems are commercially available, most of them are not extendable and so researchers cannot use them for experiments and testing new ideas. Smalltalk should prove quite a suitable tool in solving this problem [15,20]. Based on two concrete developments in Smalltalk, we discuss the appropriateness of this language to implement graphical-robot-programming environments.

Ronny M. Blomme, Jan M. Van Campenhout

Symbolic Simulation of Sensory Robot Programs

The symbolic simulation of sensory robot programs is introduced, in order to allow their systematic analysis needed to cope with difficulties arising from the fact that the program execution may result in an infinite set of possible evolutions of the robot work cell. Methods are reported for the simulation of simple sensory instructions. An experimental system for symbolic simulation is described

Vincenzo Caglioti, Marco Somalvico

Expert Systems and Multiple Criteria Decision Support

This paper aims to present the framework of a knowledge-based system which enables a re-evaluation analysis of a sample of rough solutions selected by means of an interactive multiple-criteria approach. Two case studies will be discussed, one concerning power systems expansion planning and a second one dealing with strategic telecommunication network planning. The first one will be presented in detail and the specifications of a rule-based prototype for performing the re-evaluation analysis will be discussed.

J. Clímaco, C. H. Antunes, J. P. Costa, A. G. Martins, A. T. Almeida

Signal Representation Issues in Neural-Network Modeling of Perceptual Properties

This paper discusses neural-network modeling of five perceptual processes: adaptation, scaling, similarity, absolute threshold as a function of frequency and lateral inhibition. First the need for joint-domain representations is discussed. A joint-domain representation discussed in detail is the Wigner distribution. Autocorrelations of the Wigner distribution in time and frequency are taken to view interactions of the Wigner distribution with itself. These autocorrelations are then implemented using a simple neural-network. Some insight is gained in modeling the above perceptual properties with these functions.

V. C. Georgopoulos, D. Preis

Robot Control with Procedural Expert System

In this paper the use of a Procedural Expert System (PES) for the control of a robot in a partly-structured environment is presented. Performing a task in a partly-structured environment requires a certain level of autonomy of the robot. One of the key issues needed to realize this autonomy are techniques to monitor the robot activities and to handle exceptions. The control system must be reactive to new situations. Implementation architecture and realization with PES of robot control based on the Exception Handling Model is described. The experimental results show that the event-driven mechanisms needed for the control of the robot can be adequately realized with the reactive behaviour of PES.

G. R. Meijer, T. L. Mai, E. Gaussens, L. O. Hertzberger, F. Arlabosse

Rapid Interference Detection

An algorithm for rapid interference detection between arbitrarily shaped objects is described. The approach uses a dual representation of each object, those being a wireframe and a discretized volume (octree and raycast). Interference is detected when a portion of the wireframe of one object is contained in the volume of the other. The algorithm’s speed is rooted in the speed of the point and edge inclusion tests. Tests indicate the algorithm is an order of magnitude faster than previously published work.

James S. Hemmerle, E. Levent Gürsöz, Fritz B. Prinz

Prolog Language — CADSD Package Interfacing Procedures: Applications to Heuristic Identification and Control of Dynamic Systems

The development and evaluation of different ways to interface a Prolog language with interactive CADSD environments are described. The principles of a suitable knowledge-based system solver which includes both pertinent explanations and control design software are outlined. The SERMA package has been constructed using the Turbo-Prolog shell with linkage into different CADSD environments to provide computational and graphic enhancements. The covered identification domain aspects include both a new approach defined as ‘Heuristic Identification’ of SISO-system step responses and an expert-system advisor for the model-structure determination. The control area aspect corresponds to an intelligent front-end of in-house and commercial control design packages.

Hao Nguyen-Phu

Intelligent Cooperation for Robotics

The use of Multi Agent Systems as a Distributed Artificial Intelligence paradigm for Robotics is the principal aim of our present work. In order to make different Intelligent Systems to cooperate in solving non trivial tasks, special architectures need to be designed. New functionalities are attached to each particular Agent suitable for Robotics, enabling cooperation at the Knowledge and Intelligence Levels.Each Agent is implemented by means of a set of different specialized processes. Besides the Intelligent Systems itself, a Monitor, an Acquaintance plus Self-knowledge Modules and a Communication Module, are fundamental modules that make it possible for different Agents to cooperate by exchanging competences, information and results. Which kind of knowledge to represent, either static or dynamic, involving the overall system, how to represent it, to share it, to maintain it and to infer upon it are subjects that have been adressed and several conclusions have already been drawn. An Assembly Robotics testbed is being used for that purpose, and several Intelligent Agents developed such as: An High Level Planner, a World Descriptor and an Object Identifier.Each one of these Intelligent Systems, incorporating its own reasoning capabilities (therefore considered as Scmi-autonomous Agents), has been enhanced with several modules according to the previously referred architecture. We claim Robotics to be a good example for Intelligent Systems cooperation and we intend to apply the aforementioned principles to different classes of Robotics applications.

Eugénio Oliveira, Rui F. Silva, Carlos Ramos

The Oxford Robot World Model

After reviewing the advantages of supplying a robot with a geometric model of its surroundings, we discuss the design of a modular object-oriented database to support such a model, which is being implemented as part of the Oxford Autonomous Guided Vehicle project.

Bill Triggs, Stephen Cameron

The Blackboard Architecture in Knowledge-Based Robotic Systems

Integrated knowledge-based robotic systems that employ multi-sensor feedback information require the effective treatment of a large volume of complex knowledge. To this end, several methods of knowledge organisation and exploitation have been developed. One of them is the so-called blackboard architecture, which belongs to the class of distributed problem solving architectures, and employs more than one problem solving agents. The three basic elements of the blackboard architecture are: the blackboard, the knowledge sources and the control mechanism. The blackboard plays the role of a central working memory within the system and the controller stands for the conflict resolution mechanism. The knowledge about the system (problem) is split in a number of comparatively small knowledge bases called knowledge sources and controlled through the blackboard control (scheduling) mechanism which is usually a meta-knowledge control mechanism. This mechanism is not committed to any reasoning technique (e.g. forward or backward inference chaining), but has a rather opportunistic nature (i.e., the control action is determined by the optimal decision taken at each cycle of operation). The purpose of this paper is to provide an account of the main concepts and issues of the blackboard architecture and their use in designing integrated knowledge-based intelligent robotic systems. Some particular applications of this approach will also be discussed, one of which has been designed and tested by the authors. The paper is complemented by general design issues of intelligent robotic and flexible manufacturing systems.

Spyros Tzafestas, Elpida Tzafestas

Implementation of Specific Robotic Functions Directly in Hardware

Motivated by the desire to add more intelligent functionality and enhances real-time control capability to robotic systems, many authors suggest the implementation of specific robotic functions directly in hardware. This paper presents analog and digital techniques for the computation of the direct kinematic solution (DKS) on a single chip and the implementation of a velocity and acceleration estimator based on a position history. Analog and digital circuits for the DKS have been designed on the basis of the direct kinematic equations. A velocity and acceleration estimator based on a position history is implemented with two circuits dedicated to the computation of current joint velocities and accelerations of a robot arm. The first involves a microprocessor-based system and the second consists of clock-controlled analog circuitry. The above-described units have been constructed and tested under real conditions in an experimental computer-controlled, light-weight mechanical arm.

John N. Lygouras, Basil G. Mertzios, Nicholas C. Voulgaris

The 3DP Real-Time Motion-Control Computer

Implementing advanced motion-control algorithms requires a computer which can satisfy the computational needs of the problem, a powerful development environment for programming complex and abstract ideas, and an open-architecture which gives the programmer full access to all of the system’s capabilities. Conforming to industry standards is also important for system integration with other vendors’ equipment.We have developed a system — the 3DP controller — which satisfies these requirements. The 3DP has a unique processor architecture which exploits the intrinsic 3dimensional nature of kinematic and dynamic equations. The majority of computations in advanced robot control algorithms are for solving kinematics and dynamics. By exploiting the innate 3-dimensional characteristic of the target problem the 3DP achieves execution speeds many times faster than any existing processor.The 3DP is a second generation robot control computer which improves on the ideas developed in the RIPS architecture [1]. It is an attached processor which plugs into a workstation’s VME bus. An optimizing C++ compiler lets the user program complex algorithms in a high-level object-oriented style, yet extremely efficient run-time code is produced. The user accesses the 3DP with familiar UNIX system calls, and the industry-standard VME bus allows the 3DP to communicate with other vendors’ products. Even though the user views the 3DP through UNIX, real-time program execution and external communication is not degraded by UNIX’s unpredictable response time.

Yulun Wang, Amante Mangaser, Steve Jordan, Partha Srinivasan, Steven Butner

Robot Control

A Knowledge-Based Controller for a PWR-Type Nuclear Power Plant

In this study a rule-based, fuzzy-logic regulator was developed for a PWR type nuclear power plant based on knowledge acquisition through numerical simulations and making use of a validated mathematical model of the H.B. Robinson power plant. The performance of this controller with the single objective of minimizing weighted power was found to be better in this respect when tested against that of an optimal controller using several conflicting performance criteria. The behavior of this rule-based controller was investigated also under noisy operating conditions and in the presence of drift in process variables leading to negligible degradation.

H. Levent Akin

Telepresence and Intelligent Control for a Legged Locomotion Robot

The aim of this paper is to present the actual research that is carried out in order to provide a fully operative sense of telepresence, and on-board intelligent control for a four legged locomotion robot. The paper describes the whole project showing mechanics and kinematics of the walking machine, and the teleoperation station and on-board software architecture. Special attention is given to those levels at which intelligence is necessary to improve performance or to achieve desired goals. Also, from the control point of view, the paper discusses the need and the necessary means to achieve robust and/or intelligent control under uncertain conditions posed by the operation of the legged locomotion robot in hostile and/or unstructured environments.

Manuel Armada

Robust Adaptive Control Of Robotic Manipulators

An adaptive control scheme for the control of robotic manipulators is presented. The control scheme is shown to be robust with respect to bounded disturbances and parameter variations. The adaptive controller is compared to a fixed-gain computed-torque scheme and is seen to have superior performance in the case of unknown dynamical parameters or in the case of changes in the manipulator payload.

Anastassios G. Chassiakos, Petros A. Ioannou

About Qualitative Robot Control

Among the obstacles to be overcome for reaching a full mastering of robot control, three of them seem of great importance: 1)To be able to find the optimal bridges linking the qualitative and quantitative control approaches. This point exhibits the problem of precision during a robot physical action.2)When a robot “brain” has to make a decision, to be sure that this decision is not a mistake. The reliability problem is raised with this question.3)All computations modelled into the perception and reasoning phenomena must be performed in real time. Consequently, the computational efficiency problem is brought up. The three questions are not totally independant and, up to now, no one seems able to propose satisfactory solutions. The first step must be attached to explore all qualitatives approches, knowing that quantitative approches, have been extensively studied.We present new results in four types of qualitative control. The two first ones are devoted to trajectory generation using the viability theory and fuzzy sets. The third is dedicated to an adaptative navigation control process and he last is concerned with the planning generator problem.

Ph. Coiffet, J. Zhao, J. Zhou, F. Wolinski, P. Novikoff, D. Schmit

Motion Estimation from Target Tracking

This paper describes a tracking vision system under development at the Automatic Control and Computer Engineering department (ESAII) in the Polytechnical University of Catalonia. The system consists of two recognition modules which allow the tracking of a predefined target on a multi-target environment. Binocular tracking provides, at the same time, an estimate of the relative distance which decreases the uncertainty of geometric positioning allowing a visual feedback for control purposes as well as navigation.Data association which models target motion defined by the two tracking modules, allows one to perform an oriented search based on the prediction of the motion of the target. Thus, the system is able to restrict the searching window to a defined zone of the image where foreseeable the target will be detected.The system implemented classifies the objects according to a feature vector associated with each object in the scene (area, perimeter, coordinates of the c. g, main axis of inertia with respect to the absolute reference,…). Simultaneously, the system stores a polar coding of every detected object, taking the relative reference formed by the main axis of inertia, and the perpendicular straight line through the c. g of the object.Once the user has chosen one of the targets among several which has to be tracked, its geometric features as well as its rotation invariant description (contour shape) make up its dynamic model. We call it “dynamic model” since it is updated from image to image. As a matter of fact, we can distinguish two different filtering stages. A first stage performs a rough filtering based on geometric features, while a second stage refines the classification of possible candidate targets based on the dynamic shape model with respect to their polar coding.The comparison criterion is based on a discrimination function which supplies a quantitative measurement of the divergence between the model and the potential candidates. This function implements a summation of least-square differences of 360 vectors for each candidate, coding their shapes from their relative references.This discrimination function is calculated every 40ms for all the probable targets. Thus, it is possible to define an upper threshold for this function to guarantee the recognition whenever a value below this threshold is obtained. The candidate which in every case shows the lowest function value, below the threshold, is assumed to be the target. In the mean time, the updating process is validated, and the system stores the new parameter values of both the geometric model and the shape model. This detection threshold allows a definite robustness of the system in front of sharp changes in lighting conditions, low contrast effects, low visibility, occlusions, or other unpredictable noise sources.

Juan Frau, Vicenç Llario, Joan Codina

Robust Control of Robotic Manipulators

A robotic manipulator must often be operated in an uncertain environment, e.g., picking up different payloads of different or unknown weights, accidental drop of a payload, imprecise knowledge of the robot itself, disturbances, etc. Therefore, the controller must be designed so that the manipulator will perform its tasks despite the existence of some uncertainties. In this article methods of designing stabilizing state feedback control laws are discussed based on the worst case scenario. The results are applied to the tracking problem of a robot manipulator. The second method of Lyapunov and its extensions are extensively used. Our focus is given to the controller design motivated by the game theory with conflicting agents.

F. Hamano

Automatic Determination of Quasi-Time Optimal Joint Trajectories and Adaptive Control for a Robotic Arm

In this work an original algorithm is developed to generate quasi-time optimal trajectories for each joint of a planar robot arm on the basis of the desired trajectory of its end effector, expressed in Cartesian coordinates, along with various time specifications. The algorithm takes into account the following three cases: free time for motion completion; fixed time for motion completion; motion with specified times at intermediate points. The generated reference trajectories are used in two different robot control algorithms based on the computed-torque method and the proportional-derivative feedback with feedforward control, where both methods include load adaptation.

Yorgo Istefanopulos, Luciano Casagrande

A Robust Approach for the Control of Robotic Manipulators: Predictive Control

In this paper a novel scheme, named Generalized Predictive Control (GPC) is formulated for trajectory tracking control of a two degree-of-freedom manipulator. The simulation results are presented and compared with those obtained by two other approaches, namely the conventional computed-torque method and a pole-placement algorithm. It is seen that a GPC algorithm results in a much better performance, resulting in an accurate tracking of the reference trajectory in spite of the discrepancies simulated between the robot and its model. The prediction scheme employed adds a high degree of intelligence to the system, enabling it to cope with the system-model mismatches and the nonlinearities. The paper concludes with the presentation of some results obtained from a DSP based, one degree-of-freedom experimental set up.

Okyay Kaynak

Analysis of Cooperating Robot Manipulators on a Mobile Platform

When robot manipulators are to be used in a space environment, the interactions between the manipulators and the platform upon which they are mounted increases the complexity of the dynamic model. The situation is further complicated when the manipulators must cooperate with each other to move objects or to perform assembly and disassembly operations. In this work, the dynamic equations of motion for two or more cooperating manipulators on a freely-moving mobile platform are presented. The formulation includes the full dynamic interactions from arms-to-platform and arm-tip to arm-tip and the possible translation and rotation of the platform. The equations of motion are shown to be identical in structure to the fixed-platform, cooperative manipulator dynamics. The system of cooperating robot manipulators forms a closed kinematic chain where the forces of interaction must be included in the simulation of robot and platform dynamics. The structure of the closed-chain dynamics allows the use of any solution for the open topological tree of base and manipulator links. A MATLAB simulation of two cooperating PUMA manipulators on a mobile platform is presented and displayed through SILMA Cimstation.

Steve H. Murphy, John T. Wen, George N. Saridis

On the Theory of Intelligent Machines

The Theory of Intelligent Machines has been formulated analytically. The functions of an Intelligent Machine are executed by Intelligent Controls. The Principle of Increasing Precision with Decreasing Intelligence is used to form a hierarchical structure of the control systems. The three levels of the Intelligent Control, e.g., the Organization, Coordination and Execution Levels are described as originally conceived. New designs as Neural-nets for the organization level and Petri-nets for the coordination level are included. Application to Intelligent Robots for space exploration has been used as examples in this work.

George N. Saridis

Symbolic Computation for Robot Design

This report presents the basic derivations needed to develop a CAD robot system devoted to automatic analysis and design of controlled mechanisms. The feedback laws required by sophisticated robot controllers are directly derived from the Lagrangian of mechanical systems by means of generating rules. The dynamical model is not explicitly needed in this derivation which only requires partial derivatives and matrix calculus. In addition it is shown how the dynamical model can be computed from the control laws so that dynamics of controlled and uncontrolled mechanisms can be derived in the same framework. Finally, the sampling rate to use in computerized servo-mechanisms is defined from a simple rule.

C. Vibet

Neural Net System for Adaptive Robot Control

Simple relationships can be employed in the control of robot manipulator links. In this paper the use of back propagation nets in configuring such controllers is considered. The ability to learn, firstly via training and secondly by experience, is made use of to provide an adaptive structure. An application example is considered in order to show how such controllers can be implemented.

Kevin Warwick


On-line Scheduling and Planning of Robots and Presses

The work reported in this paper deals with the application of Expert Systems (ES) approach to the scheduling and planning of a set of presses and robot feeders in an automobile factory. This work has been usually done by experts in the field with a sound knowledge in all the related topics: Processes, requirements, priorities, constraints, options and decision timing. One of the most important constraints is the response time. So, time constraints have been considered in the design and implementation of the ES solution. In the first section of this paper, a description of the process is presented, pointing out the features of the application, the structure of the ES, and the temporal constraints. In the next section, an overview of the model used to represent the temporal knowledge will show the possibilities of this approach. Finally, the main ideas of this application are discussed, and preliminary results are shown.

A. Crespo, F. Barber, P. Albertos

Knowledge-Based Process Planning, Scheduling and Error Recovery in Robotized Assembly

The goal of current researches in robotized assembly is to increase both productivity and flexibility. In this way the trend is now to use two or more robots working together on the same assembly, sharing a part of their working area and a number of tools.This paper describes a practical approach to the control system of a two-robot assembly cell. This system consists of four main functions: process planning, scheduling, supervising and diagnosing, and error recovery.The planning phase generates the assembly plans from geometrical and topological descriptions of the product. From a chosen assembly plan and from a rough description of the production means, the scheduling module answers the question: “Who has to do what, when and how?” Finally, when an error has been detected, the error recovery module puts the cell in such a state that the assembly can restart.In this paper, the specific knowledge and the common sense expertise are specified for each module.

A. Delchambre, P. Gaspart, D. Coupez, P. R. Glibert

Complete Garment Assembly Using Robots

This paper describes the evolution of the problem of robotic based fabric handling as seen by the Robotics Research Unit at Hull University. The third phase of the associated project, following success in ply separation and robotic sewing, is to produce a prototype robotic assembly line for a complete garment. The line is designed for assembling mens Y-fronts and ladies briefs. The project is subdivided into two dimensional component based assembly, three dimensional folding and closing operations, and a supervisory computer system to aid communication and reconfiguration.The manual operations for constructing these garments are mentioned in this paper, outlining the handling difficulties and the required system goals in terms of cost and performance. The various robotic handling solutions are described in detail, together with an assessment of their performance. Conclusions include the implications of the techniques used in this assembly line on solving the problems associated with products of a similar nature.

I. Gibson, P. M. Taylor, A. J. Wilkinson, G. S. Palmer, M. B. Gunner

A Model-Based Expert System for the Diagnosis of Faults in a Robot System for Cleaning Castings

This contribution presents an Expert System which has been developed within a joint project of the University of Karlsruhe and the Nuclear Research Center Karlsruhe.The long term goal of this cooperation is to develop a generic tool for the diagnosis of faults of technical systems. In order to make such a tool generic enough to be easily adaptable to new applications, a deep-reasoning approach has been developed. A prototype of the system has been implemented using the Knowledge Craft (TM) environment The prototype has been tested with industrial robots of UNIMATION, Inc. and is currently being applied to the much more complex COMETOS system for the cleaning of large castings consisting of several robots, end—effectors, sensors and different controllers. Both applications are described.

Thomas Hübner, Klaus Hörmann

Genghis: An Intelligent Autonomous Mobile Robot

This paper describes the control system of Genghis, a small robot that can track people over rough ground. Genghis behaviors are made of simpler behaviors, each of which is composed of combinatorial logic. Some of these behaviors are connected to sensors and some are connected to motors. All behaviors have a reset, some have memory. All behaviors can run at the same time, while higher behaviors can overide lower behaviors.

Mark M. Simpson, Tim Jordanides

Benefits of Expert Robots: Intelligence vs Skill

The increasing costs of recruitment and shortages of skilled labour have been the main driving forces behind the recent advances in Automation Technology. Automated systems have brought significant economic benefit through consistency, repeatability, control and improved quality in production within the manufacturing, aerospace and food industries. During the next decade the scale of this benefit to Europe will be even greater as research and technological developments mature into advanced robotic and automated systems for factories currently employing the declining population of skilled labour.

K. Khodabandehloo

Evaluation of Applicative Assembly Design, Supported by Expert Systems

This paper deals with the application of an expert system for the design of an assem bly process with robots. The system to be assembled is an electric motor, composed of seven elements. The design process is composed of: geometrical analysis, compatibility of components with robot tools, design of new tools, preparation of logical structures of assembly process, realization of the expert system able to determine the optimal assembly sequence, evaluation of technical characters for different sequences, the final choice, connection with the robot, and actuation of the robot. A second part deals with the expert systems’economic evaluation, in regard to productivity and economic results, until now limited to the realized cases.

Alberto Rovetta, Elio Zamponi, Vitor F. Romano, Remo Sala

Reports of Group Meetings

Group Findings: Vision and Image Analysis as Applied to Intelligent Modules

There was a common interest in vision systems among the participants with varying levels of expertise ranging from the total inexperienced observer who wanted to learn, to those who had a number of years experience in applying machine vision technology. A full list of participants was not possible, but a partial list was made, comprising the following personnel:

Ian Gibson

Computer-Aided Process Planning in Robotic-Based CIM Systems

This paper summarizes the discussions of a working group on computer-aided-process-planning (CAPP) in robotic-based CIM systems. The croup consisted of scientists and engineers with diverse backgrounds and expertise in robotics and manufacturing. The discussion focused on defining terminology and outlining the problem scope in order to bring such a diverse group to a common level of understanding.

Lee E. Weiss

Current Issues and Future Developments of Intelligent Robots

K. Khodabandehloo

ASI Unusual Robot Applications

K. Khodabandehloo

Industrial Needs and Applications

K. Khodabandehloo

Integrating Diverse Knowledge: New Ideas in Knowledge Processing

ASI Discussion Group on Advanced Knowledge Processing

There was considerable interest at the ASI in Knowledge Processing Systems (KPS) for intelligent robotics and integrated manufacturing, so an informal discussion group was formed to consider some of the issues raised by current attempts to produce such systems. The group was small but most of its members were actively involved in KPS applications and the lively discussion was continued after the meeting by correspondence.In order to achieve its goals, an intelligent robot or manufacturing system must combine large quantities of diverse knowledge (ie structured factual and procedural information) into a useable whole. The key problem in these domains is thus the integration of diverse knowledge. The group considered several interconnected topics in relation to this: System architecture and control strategies.Knowledge representation and accessing.Knowledge application strategies.

Bill Triggs, H. L. Akin, A. Akmehmet, G. Honderd, K. Khodabandehloo, W. Kropatsch, E. Oliveira

Intelligent Control

This article summarizes the presentations and discussions from the group meeting on Intelligent Control which was held as a part of the NATO Advanced Study Institute on Expert Systems and Robotics, Corfu, Greece, July 1990. The meeting consisted of the short presentations given by Professors G. N. Saridis, S. Tzafestas and G. Honderd respectively on the definition and limitations of Intelligent Control, need for clear definition of intelligent control and need of Intelligent Control from the industry point of view, followed by the discussions by all the participants.

Fumio Hamano


Weitere Informationen