Skip to main content

Über dieses Buch

This volume contains papers presented at the NATO Advanced Research Workshop (ARW) on "Sensors and Sensory Systems for Advanced Robots", which was held in Maratea, Italy, during the week Apri I 28 - May 3, 1986. Participants in the ARW, who came from eleven NATO and two non-NATO countries, represented an i nternat i ona I assortment of d i st i ngu i shed research centers in industry, government and academia. Purpose of the Workshop was to rev i ew the state of the art of sensing for advanced robots, to discuss basic concepts and new ideas on the use of sensors for robot control and to provide recommendations for future research in this area, There IS an almost unanimous consensus among invest i gators in the fie I d of robot i cs that the add i t i on of sensory capabi I ities represents the "natural" evolution of present industrial robots, as wei I as the necessary premise to the development of advanced robots for nonindustrial app I i cat ions. However, a number of conceptua I and techn i ca I problems sti I I challenge the practical implementation and widespread appl ication of sensor-based robot control techn i ques. Cruc i a I among those prob I ems is the ava i lab iii ty of adequate sensors.



The Biological Model: The Human Body and Its Sensory Systems


The Central Nervous System as a Low and High Level Control System

The division of the central nervous system into high and low level control systems has long been recognized. How these two systems are interconnected and how they influence each other is one of the great mysteries of modern science. Recent attempts to produce intelligent behavior in robots and computer integrated manufacturing systems have produced insights as to how high level goals can be decomposed into low level actions, and how knowledge about the environment can be acquired, stored, and accessed by task decomposition processes to produce sensory-interactive goal directed behavior.
James S. Albus

Proprioceptive Feedback for Sensory-Motor Control

Among the natural sensors, which provide proprioceptive feedback during movement, the muscle spindle has attracted most interest since the sensitivity of spindle afferents to length variations of the host muscle is subjected to efferent control by fusimotor neurones (mostly γ-motoneurones). These are themselves controlled from numerous centres of the nervous system. Functionally, the perhaps most important feature of γ-action is that the fusimotor system, with its static (γS, sensitivity reducing) and dynamic (γD, sensitivity enhancing) components, possesses the potential of providing flexible gain (or sensitivity) control of spindle feedback. One of the most relevant pending issues in the field of peripheral control of movement is whether, and under which circumstances, this potential is used during physiological motor performance.
This paper gives a brief overview of important functional properties of the sensory spindle afferents and of efferent fusimotor action and it addresses the question of fusimotor function during natural movement. It is pointed out, that strategies of fusimotor action, which have been encountered in reduced laboratory preparations, appear not be be adhered to during normal motor performance. An alternative concept of fusimotor function, the notion of fusimotor set, is presented. According to this, key features of fusmotor action are
  • largely tonic γS and γD firing patterns (even during rhythmic movements, featuring rhythmic skeletomotor α-activity), suggesting independent central control of the skeletomotor and fusimotor systems,
  • a dramatic switch from predominant and low-key static action (during routine motor performance) to predominant and powerful dynamic action (during unfamiliar motor tasks),
  • gradual resetting of fusimotor drive of a given type during periods of motor adjustment.
Thus, proprioceptive feedback from spindle afferents is not invariant. It appears to be adapted to specific requirements of motor tasks, conceivably to optimize motor control.
Manuel Hulliger

Physiology and Psychophysics in Taste and Smell

The ability to recognize chemical stimuli in the external environment is a feature of living organisms ranging from bacteria to man. It has played a crucial role in the search for food, detection of adverse environments, and sexual behaviour. The transduction process of chemoreception in both lower organisms and higher vertebrates consists of binding of chemical stimuli to a receptor membrane, giving rise to a receptor generator potential. This potential either triggers a chemotactic response in lower organisms or is transmitted in the form of neural impulses to signal processing centres in higher animals. In vertebrates, chemical stimuli in external environments are received at gustatory and olfactory receptor cells. These cells detect chemicals, encode information about the intensity, duration and quality and transmit it to higher processing centres. This chapter discusses some aspects of current research in the physiology and psychophysics of taste and smell in vertebrates.
K. C. Persaud, J. A. DeSimone, G. L. Heck

The Physiology and Psychophysics of Touch

We have selected material for this presentation by structuring the discussion around a series of critical questions that might be asked with a view toward modelling a biological tactile sensing system.
S. J. Lederman, R. A. Browse

The Artificial Counterpart: Robots and Their Sensory Functions. General Remarks


Steps toward making robots see

This paper reports on recent progress in Computer Vision by the Oxford Robotics Research Group. We discuss in particular: edge and corner finding; shape from contour; parallel algorithms for computing shape representations; parallel architectures for computer vision; and the application of truth maintenance systems to recognise variable geometry objects in cluttered images. Model-based vision and data-directed vision are discussed as extreme cases of architectures for vision systems.
Michael Brady

An Overview of Local Environment Sensing in Robotics Applications

This paper presents an overview of the basic technical solutions for local environment sensing, and gives some examples of realizations. A first section sets the problem of local sensing through three concepts: relativity of the informations, necessity of compactness, fastness of the related control loops. Section 2 gives a comparative presentation of the possible local sensors and describes contact, magnetic and ultrasonic sensors. Section 3 is devoted to optical sensors, including local vision and optoelectronic systems: triangulation, phase shift and reflectance sensors. Section 4 deals with multisensory systems and local control. The paper ends with the presentation of some realizations and gives some remarks about the future of local environment sensing.
B. Espiau

Force and Tactile Sensing for Robots

This paper deals with methods and techniques for sensing and controlling contact forces originated by the physical interaction of a robot with the environment.
Contact sensing includes force and tactile sensing. A distinction between these two sensing modalities is that force sensing refers primarily to the measurement of the resultant mechanical effects of contact, while tactile sensing involves the detection of a wide range of local parameters (physical and chemical) affected by contact.
This paper is comprised of a section in which force sensing techniques and devices usable in different classes of robots are discussed, and a second section dealing with true tactile sensing. In each section the motivations and the general problems encountered in each sensing modality are analyzed first, then the state of the art of related sensing technologies is surveyed.
In a third section, a coherent approach is presented to the problem of replicating human tactile sensing capabilities in an artificial robotic system. The advantage of this approach is to allow a comprehensive analysis of the sensory, motor, and control components of an advanced robot system, as well as to illustrate the roles of force and tactile sensing, and the importance of their coordination, in artificial tactile perception.
P. Dario, M. Bergamasco, A. Fiorillo

Physics of Transduction and Transducing Techniques


Basic Transduction Mechanisms and Techniques

This paper presents a scheme which can be used in classifying solid-state-sensors. The scheme is based on the First Law of Informatics which relates information to mass and energy. The various kinds of measurands are discussed and grouped into six signal domains. The main physical and chemical effects that can be used to convert non-electrical signals into electrical ones are presented. Finally, a three-dimensional diagram, the “Sensor Cube”, in which all effects can be placed is introduced.
S. Middelhoek, A. C. Hoogerwerf

Analogs of Biological Tissues for Mechanoelectrical Transduction: Tactile Sensors and Muscle-Like Actuators

In this article the authors report their current attempts toward the development of new “skin-like” tactile sensors and “muscle-like” linear actuators potentially useful in the design of dexterous end effectors. The underlying design philosophy resides on mimicking electromechanical conversion properties of biological tissue making use of synthetic piezoelectric polymers or polyelectrolyte gels. A brief introduction is also given to the physical mechanisms which govern mechanical to electrical transduction in polymeric systems. It is a belief of the authors that substantial progress in the development of sophisticated tactile sensors and artificial muscles can be obtained by resorting to a “molecular bionics” approach.
Danilo De Rossi, Claudio Domenici, Piero Chiarelli

Solid State Transducers

This paper reviews the development of solid state sensors and actuators during the last decade. Both solid state physical and chemical sensors suitable for use in the measurement and control of manufacturing processes will be discussed. The principle, the device structure and the performance of physical transducers for temperature, pressure, displacement, acceleration, flow, display, fluid injection and controlled fluid flow valves is summarized. Chemical sensors for humidity, combustive gases, ionic concentration in electrolytes and large molecule concentration are also outlined for their potential applications in automated manufacturing.
The technology of micromachining, one of the bases for solid state transducers, is discussed with examples given to illustrate its capabilities and remaining problems.
The possibility of integrating signal processing circuits on or near the transducer chip to fabricate “intelligent sensors” is presented along with the functions that can be incorporated on the transducers.
Future trends in solid state transducer research and the possibility of beneficial collaboration between users and designers, as well as between material scientists, technologists, device designers and packaging engineers, is discussed.
Wen H. Ko

Fibre Optic Sensors in Advanced Robotic Systems

The apparent demands from advanced robotics for sensory systems with a complexity approaching that of a human being inevitably lead to the conclusion that a novel approach to sensor technology will be required. Fibre optic sensors are currently in their technological infancy though have demonstrated impressive performance in the laboratory. This indicates that the technology could prove to be a viable approach for such sensory systems. In this paper the basic principles of fibre optic sensors are briefly reviewed, some current examples are outlined and some possiblities for the way forward are discussed.
Brian Culshaw



Models of Errors and Mistakes in Machine Perception: First Results for Computer Vision Range Measurements

This paper is a first step toward integrating sensor measurements of distance. Its major contribution is to identify and present qualitative models for the errors and mistakes introduced in three particular computer vision distance measurements: range from focus, range from point-based stereo, and range from line-based stereo.
These range measurement techniques are presented as computations, and their dominant sources of error are analyzed qualitatively. We propose to quantify the underlying models for these three range estimation techniques by deriving approximate confidence procedures for the intrinsic parameters and functions which characterize each technique.
Ruzena Bajcsy, Eric Krotkov, Max Mintz

Holographic Techniques

Robots are flexible handling systems. To use this flexibility best, the robots should get more intelligence, especially with recognizing sensors. One necessary form of recognition is the identification of objects in the three-dimensional space, or more precisely: there must be a projection of the object to be handled — or not — into the “brain” of the robot, fig. 1.
W. Jüptner

Electrical Proximity Sensors

The subject of this paper is the use of electrical devices for proximity sensing. A general discussion is made about the concept of multisensory “perception” by the robot, and about the interest of proximity sensing for reducing the complexity of this problem. The main characteristics of sensors using the electrical field, the magnetic field, or electromagnetic waves are then discussed, from general physical arguments. Indications are given on actual sensor design and on available devices. An exemple of the characterisation of a multisensory system is given for an eddy current array developped in our laboratory.
H. Clergeot, D. Placko, J. M. Detriche

Ultrasonic Sensors

This paper reviews the field of ultrasonic sensors for use in robotics with emphasis on sensors based on piezoelectric materials for use in air and in water. A detailed discussion is given of demands to sensors used for range and proximity measurements as well as of factors influencing the accuracy of these measurements. The structure of the ultrasonic field of a piston source is emphasized and the influence of sensor backing and load as well as bond lines on characteristic features of the ultrasonic field is discussed. The use of focused sources and aperture shading is evaluated and ultrasonic sensor calibration methods are presented. Finally, several ultrasonic sensor types for use in air and in water are presented including bimorph sensors and their characteristics.
Leif Bjørnø

Acoustic Imaging in Three Dimensions

An algorithm and method of implementation are described for obtaining high-resolution, three-dimensional acoustic images of simple objects in air using only a small number of transducer positions. Each transducer position provides a single large bandwidth pulse-echo waveform along a specified incident direction. Probabilistic estimation procedures using the full waveforms are used to reconstruct a three-dimensional image of the objects in the observed volume. Prior information about the physics of the scattering process, the statistical nature of the noise and the acoustic reflectivity of the target object are used to obtain resolution of the order of the smallest wavelength present in the acoustic pulses. Results of both synthetic and experimental tests of the algorithm are presented, as well as a brief description of the experimental apparatus and the algorithm used.
J. F. Martin, K. Marsh, J. M. Richardson, G. Rivera

Gas Sensors: Towards an Artificial Nose

The design of an artificial nose requires (a) a knowledge of the “olfactory language”, i.e. the molecular parameters of the odorants that are measured by specific receptors in the nasal mucosa; (b) the availability of gas sensors, specific, reversible and sensitive, able to convert chemical informations into electric signals.
For what concerns the understanding of odour recognition at the molecular level, it is known that specific receptor proteins, present in the olfactory mucosa, discriminate the various molecular structures on the basis of size, shape and position of functional groups: in this process the hydrocarbon part of the molecule is often very critical.
Gas sensors, based on polymers of aromatic and heteroaromatic molecules, that exhibit electrical conducting properties, can be used as specific odour detectors. Their electrical resistance varies upon interaction with several substances in the gas phase, with a specificity related both to the presence of functional groups and to the shape of the hydrocarbon part of the odorant molecule.
This characteristic, together with fast responses and high stability, make such polymers suitable for designing an artificial nose. Moreover, their organic nature would allow the synthesis of new polymers tailored for detecting particular gases or for spscific needs.
Paolo Pelosi, Krishna Persaud

Development and Application of Humidity Sensors

After a short discussion on the application-fields of humidity sensors and the related requirements on specifications, a review of measurement principles Is given together with some general characteristics of these methods. Two examples are worked out: absorption sensors and dew-point sensors. A review of the state-of-the-art is given as well as present trends in research on such sensors.
P. P. L. Regtien

Development and Application of Chemical Sensors in Liquids

In order to measure the chemical constituents of a liquid, such as various ionic concentrations, dissolved gases, biochemical species etc. by electronic means, such as various types of recording equipment, a large variety of electrodes has been developed. Focussing attention on these electrodes, there will always be an interface, or a system of interfaces, in which the “ionic world” communicates with the “electronic world”.
P. Bergveld

Integration of Sensory Systems


Integration of robot sensory systems

The views expressed in this paper have been formed as a result of attempts to fabricate array tactile sensors. Whilst an array of 20 × 15 elements of 1 mm pitch has been successfully fabricated, it is now clear that shape recognition, which is all that this sensor is capable of, is only a limited facet of tactile sensing. It is, therefore, valuable to examine the nature of tactile sensing in human beings.
B. V. Jayawant

Multisensory Feedback including Cooperative Robots

For on-line sensory feedback in robotics several major prerequisites are necessary:
the design of small, cheap gripper sensors with decentralized signal processing
the availability of systematic procedures for generating adequate feedback structures
tight man-machine interfaces
This paper outlines the recent sensory developments in the DFVLR robotics lab, as are different force-torque sensors, laser range finders, inductive sensors and sensor balls for robot and 3D-computer-grafic control. With the example of our preferred multisensory arrangement implying vision, range sensing, force-torque and speech, the fine-motion planning and path generation techniques as developed in our lab are discussed. The special case of a two-arm cooperative robot using two force-torque-sensors is treated in more detail. An outlook to space robotics in terms of an experiment proposal is given.
G. Hirzinger, J. Dietrich

Active Vision Integration of Fixed and Mobile Cameras

The integration of vision and movement is investigated. Active control of egomotion is a powerful source of constraining perceptual processes. Stereometric strategies related to movement are compared: triangulation strategies (binocular and kinetic) and volumetric strategy (based on occluding contours). Volumetric representations are considered as candidate structures for accumulating a flow of stereometric data. Software tools and experimental results are presented.
P. Morasso, G. Sandini, M. Tistarelli

Sensors for Mobile Robots

Mobile robots represent a new generation of automation. Technologies developed for industrial robots in the field of Production Automation are still being generalized for application outside the factories. Mobility is a characteristic feature for this kind of robots assigned for a lot of different applications extending from space to underwater purpose, from construction to agriculture and from mining to fire fighting.
G. Drunk

Multimedia Sensory Systems for the Solution of the Error Recovery Problem

The advanced role of sensory systems is very important in intelligent assembly robots.
Two special-purpose techniques are developed for sensory perception and planning, by exploiting the particularity of the problem they are meant to deal with:
to provide a robot with both the capability in understanding the causes of errors in its activity and the capability in correcting the error situation and in recovering its fully operational functionality.
A robot, while is executing a working cycle, may be stopped within an assembly process by an error caused by a defective object, by a wrongly positioned object, by the absence of an object, or by an unpredicted collision between objects.
Therefore, when such unpredicted (and undesired) events happen, the correct execution of the cycle may be compromised or even stopped. Sensors can then be utilized to extract knowledge about the actual (error) situation. This knowledge can then be employed in order to plan the correction of the occurred error.
In the sequel, we shall call this set of activities as the activity of error recovery.
The purpose of this paper is to show how the integration of multimedia sensory data may become functional for the solution of the automatic error recovery probIem.
Before a planning a strategy intended to error recovery, the elements to be considered within the planning problem have first to be identified. These elements are:
  • the starting state (in this case the error state);
  • the target state, which has to be determined according to the task, that is assigned to the robot.
Therefore a comprehensive correcting activity can be arranged along three phases:
the deduction of the task, starting from the assigned program and from possible further information;
the detection of the error situation, through the use of sensors;
planning a correcting strategy, whose execution allows the correct accomplishment of the robot task, in spite of the error that has occurred.
Our approach is based on the the adoption of a three-modules architecture.
A first (off-line) module, starting from the program assigned to the robot and using some knowledge about physical and geometrical features of the objects, makes a knowledge based task deduction, via a simulation of the program execution (“dynamic” action’s analysis), in order to provide a wide representation of the desired evolving world state (in terms of positions, contacts, and constraints among the objects).
A second (on-line) module, after an error has occurred, plans a dynamic sequence of sensory detections, in order to acquire a sufficient amount of knowledge to construct a useful description of the actual (error) situation. The attribute “dynamic” refers here to the fact that the next sensory detection of the sequence depends on the current world model, that has been constructed exploiting the result of the current sensory detection.
A third (on-line) module plans a correcting strategy, using as starting state the description of the error state (provided by the second module), and using as target state one of the desired evolving states (provided by the task deduction module).
This approach allows to reduce the weaknesses indicated above, but has the drawback of a supplementary processing charge, especially during the task deduction activity.
However, the fact that the task deduction is executed off-line reduces the importance of this shortcoming.
V. Caglioti, R. Simino, M. Somalvico

Pattern Recognition in Multidimensional Perception: Robots and Humans

We discuss several aspects of the theory of multidimensional human and robot perception which is important for developing robotic sensory capabilities beyond the ones currently in use, i.e. beyond black & white vision and force/tactile sensing. Understanding Multidimensional Perception (MDP) is necessary for developing sensory capabilities for color vision, for general cases of force/torque/tactile sensing, and for thermal/chemical sensing. In this paper we describe a new method of pattern recognition in robotic MDP:Class Connectivity Analysis (CCA). We also describe a new effect in human MDP: the Color Void effect. Finally, we propose a new design criterion for MDP in robots. The CCA method allows fast inspection of complex images, in particular textured images and random patterns. The Color Void effect clarifies the mechanism of human pattern recognition in MDP and suggests a bionic design criterion for MDP. This criterion could allow the description and recognition of patterns in sensory fields with arbitrary time-dependent signals.
Gerardo Beni, Susan Hackwood


Weitere Informationen