bertnoro03 fiorini 1 - PRISMA Lab · Paolo Fiorini Department of Infor mat i c s Uni ver s i t y of...

1

Laboratory for Teleoperation and Intelligent Robotics

Advanced Telemanipulation

Applications

Paolo Fiorini

Department of Informatics

University of Verona

Bertinoro July 16, 2003

Bertinoro July 16, 2003 2/61Laboratory for Teleoperation and Intelligent Robotics


Summary

• Introduction

• A biased history of teleoperation devices

• Joystick survey and analysis

• The Universal Force Reflecting Hand Controller (FRHC)

• Teleoperation control strategies

• Verification of Teleoperation Systems

• Conclusions



The Problem

Provide enough perception and command capabilities to a

human operator so that he/she can control a remote robotic

device as if the operation were carried out by hand, i.e.

achieve complete Telepresence and Telecontrol.

2



� Remote control and maintenance

� Teleoperation and telepresence

� Computer assisted surgery

� Training and evaluation

� Operations in high risk

environments

� Advanced Human computer interfaces

� Rehabilitation and assistance

Applications



� Trade off between stability and perception

� Compensation of communication time delay

� Low inertia, low friction, high torque mechanical design

� High performance actuators and sensors

� Realistic models of tasks and environment

� Effective training and validation procedures

� Integration of different sensory channels

� Design and cognitive load validation

A Few Technical Challenges



Summary

• Introduction






• Conclusions

3



A Biased view of Past Work

Used in the 1940-1950 to handle radioactive materials

The Argonne devices




NASA early devices

� Developed in NASA in

the ‘70

� Identical master and slave

� Master exoskeleton

� Capable of force

reflection

� Without on board

computing




JPL CURV Arm (1981)

4



A Biased view of Past WorkJPL Universal System (1988)




JPL Master Station (1989)

� Universal

master with

multi-sensor

interface

� Tested

extensively with

differente types

of teloperation

architectures




JPL Slave System (1990)

Remote robot

system planned

for repair on

Space Station

5




JPL-Jau force reflecting robotic hand (1992)




JPL Microsurgery system (1995)



A Biased view of Past WorkTeleoperation for Robonaut (1996)

6



A Biased view of Past WorkVisual teleoperation for Robonaut (1999)




Solar Max Repair Simulation

at NASA-JPL




Exoskeleton teleoperation at

NASA-JPL

7



Summary

• Introduction






• Conclusions



Joystick Components• A Hand Controller (HC) consists, generally, of two separate

components: the controller and the handle.

• Then the design must address the control input mode and the

control techniques.

• The combination of these four elements determines the

performance of the Master station of a teleoperation system.



Joystick Requirements

The design of a Hand Controller include at least the following

features:

• The handle, must have sufficient stimulus-response compatibility,

• The handle must not be fatiguing,

• The HC should incorporate force feedback,

• The HC should have proportional position feedback,

• The handle should be compatible with the controller structure,

• The handle should be usable by the 5-95% operators.

8



Human Grasp Capabilities



Feature Definitions

Stimulus-response compatibility means that the operator's gripping

action should have a natural correspondence to the grasping action

of the slave. This is essential for good control and to prevent

operator confusion.

One of the most significant cause of errors in telemanipulation is

operator's fatigue. The endurance of an operator to maintain a given

muscular force is related to the magnitude of the force and the time

over which it must be exerted. Thus when the operator is required

to exert a grasp force over an extended period, the force should be

well below the individual's maximum force capability.



Human Hand Variability

9



Feature Definitions

• Human variability is addressed by requiring that the handle be

usable by the 5% female to the 95 % male user (adjustments may be

required).

• Some form of force-scaling greater than 1 from the slave to the

controller handle, must be included in the design. This scaling is due

to the fact that none of the typical forces exerted in the human

grasps cover the complete range of required forces.

• It must be noted that the operator's grip strength and the handle

controllability, depend not only on the physical attributes of the

operator, but also on other design elements, such as handle width,

height, contour, texture and grip location.



Handle Mounting Considerations



Feature Definitions

The attachment of the handle to the hand controller must be compatible

with the intended use of the system and its surrounding environment.

• In the first design, the handle on top of HC structure, makes the operator

support the weight of the HC, thus increasing the potential for slippage of

the handle, and for early fatigue, since he must squeeze the handle harder.

• The second solution, handle above HC, results in a very evident

interference problem, since the operator's forearm is below the HC

structure, and the upper arm is above it.

• Finally the third design, with the HC structure in front of the operator,

forces the loss of valuable space in the operator control room.

10



Handle Design Guidelines• The handle must supply kinesthetic and force feedback,

• the handle shall incorporate:

• grip lock/release switch, secondary function switches, and dead-man switch.

• the handle shall accommodate a full range of operators,

• gripping action shall have direct proportional correspondence to the grasping action of

the slave,

• handle configuration shall be compatible with the controller structure and will allow a

full range of movement,

• switches and feedback mechanisms shall be designed and placed to allow direct and

uncumbersome actuation without regripping action by the operator,

• pressure required to activate switches and gripper shall not approach the requirements of

the least capable operator within 25\%,

• switches shall be designed to prevent accidental activation,

• handle shall be lightweight.



Handle Design EvaluationHandle design should be evaluated according to four main categories:

• Engineering Development -- (i) design simplicity, (ii) difficulty of implementation, (iii)

availability of the required technology, and (iv) cost.

• Controllability -- (i) stimulus-response compatibility, i.e. the degree to which a design

approaches, or improves, on the industry standard compatibility, (ii) cross coupling

between arm motion/forces and the grasp considers the coupling between the desired

motion, or forces, of the arm and thedesired motion, or forces, of the gripper.

• Human-Handle Interaction -- (i) secondary function control, considering the placement

of the secondary switches form the stand point of activating a given function; (ii) force-

feedback ratio, considering the extent of the scaling of the remote forces; (iii) kinesthetic

feedback, considering the motion range of the device commanding the opening of the

remote gripper; and (iv) potential for accidental activation.

• Human Limitation -- (i) endurance capacity, considering the relative duration, as

compared to other designs, that allows operation without fatigue and/or stress; and (ii)

operator accommodation, considering the extend to which a given design can

accommodate different operators.



Input Devices Switch Control. Consist of simple spring-centered, three position (-, off, +), discrete

action switches (toggle, push/pull, or slide), where each switch is assigned to either a

particular manipulator joint or spatial degree-of-freedom of the end-effector.

Potentiometer Controls. Are used for proportional control inputs. They can be either

force-operated (e.g. spring centered), or displacement operated. Typically, each

potentiometer is assigned to one manipulator joint or a spatial degree-of-freedom of the

end-effector.

Isotonic Joy-stick Controller. Is a position operated fixed-force (isotonic) device used to

control two or more degrees-of-freedom with a single hand, from within a limited control

volume. The controller output does not correspond to the forces applied by the operator

and the control lever remains in the last position set (the joystick usually maintains a set

position by virtue of sliding friction). A trackball is a well-known example of an isotonic

joystick.

Isometric Joystick Controller. Is a force operated minimal-displacement (isometric)

device used to control two or more degrees-of-freedom with a single hand. The controller

output corresponds directly to the forces applied by the operator, and drop to zero unless

manual force is maintained.

11



Input Devices • Proportional Joystick Controller. Is single-handed, two or more dof device with a limited

operational volume in which the displacement is a function of the force applied by the

operator (F=Kx). the controller output corresponds directly to the displacement of the

device.

• Hybrid Joystick Controller. Is a controller composed of isotonic, isometric and

proportional elements (which are mutually exclusive for a given dof), used to control two

or more dof from within a limited volume with asingle hand. There are two basic

implementation philosophies: concurrent (some dof isotonic and others isometric or

proportional) and sequential (initially act as isotonic and then switches to isometric).

• Replica Controller. Is a device which has the same geometric configuration of the

controlled manipulator, but which is built on a different scale. Hence, there is a direct

correspondence between the joint movement of the replica and the teleoperated arm.

• Master-Slave Controller. Is a device which has the same geometric configuration and

physical dimensions as the controlled manipulator, as well as a direct 1:1 spatial

correspondence between the joint motion of the master and the slave. generally, master-

slave systems are bilateral, but may also be unilateral. In master-slave system, typically

the master is mechanically linked to the slave.



Input Devices • Anthropomorphic Controller. Is a device which derives the manipulator control signals

from the configuration of the human arm. The device may or may not have a geometric

correspondence with the controlled manipulator. However, when a geometric

correspondence does exist, anthropomorphic controllers have the added advantage that

they provide direct configuration feedback to the operator through his arm.

• Non-Geometric Analogic Controller. Is a device which does not have the same

geometric configuration as the controlled manipulator, but which maintains joint-to-joint

or spatial correspondence between the controller and the slave. They are used when the

general characteristics of a master-slave manipulator are desired, but where overriding

design constraints preclude the use of that option.

• Universal Force-Reflecting Hand Controller. Is a six dof's control device which, through

computational transformations, is capable of controlling the end-effector of any geometric

dissimilar manipulator. The device is essentially a large volume joystick, except that it

can be endowed, through computation, with isotonic, isometric, proportional and hybrid

characteristics, without modifying the device itself.

• Universal Floating-Handle Controller. Is a completely non-geometric six dof's control

device, without joints or linkages, which is used for controlling the slave end-effector in

hand-referenced control.



Grip and Trigger Concepts

12



Summary

• Introduction






• Conclusions



JPL-Stanford Hand Controller



JPL-Stanford Hand Controller

Low Backlash. A cable drive system was used to virtually eliminate backlash in the

drive train.

Low Friction. In order to keep the forces required to back drive each axis less than

10% of the maximum output force, a system of pulleys was used. Precision bearing

minimize friction, and large radius pulley to reduce friction and cable bending.

Low Effective Inertia. A cable tensioning system permits the actuator for the last

four axes of motion to be stationary with respect to the hand grip. The effective

mass of the grip is thus kept less than 1 Kg and the inertia less than 1Kg-cm2.

Neutral balance. A counter balancing mechanism is included as part of the cable

tensioning system, to eliminate the need for actively calculating and applying

torques to offset gravity forces.

Simple Force and Position Transformation. A simple kinematic design with

intersecting axes has been used to minimize the computational burden of

transforming forces and positions to and from world or hand coordinates and joint

coordinates.

13



JPL-Stanford HC Kinematics



JPL-Stanford HC Details



Summary

• Introduction






• Conclusions

14



Teloperation Control Strategies

• Rate control

• Unilateral position control:

• Bilateral position control:

• Operator aiding control:

• Filtering,

• Scaling,

• Rereferencing,

• Controller,

• Control Coordinates,

• Motion compensation,

• Motion constraints,




Rate control

Direct rate control occurs when the controller output is relayed directly

to the manipulator servos, where it is interpreted as an actuator velocity

command. The controller dof's typically have a one-to-one

correspondence with the manipulator dof. The commanded velocities

can be either preset or continuously variable, depending on the

controller used.

Advantages: A small controller motion can cover large workspace

accurately. The accuracy of manipulator positioning does not depend on

joint resolution. Simple implementation.

Disadvantages: Operator must mentally coordinate the input commands

to obtain the desired end effector motion. Generally it is not compatible

with force feedback. The end-effector location must be obtained

visually from the remote images, or via mental integration.




Resolved Rate Control

The controller output is interpreted by a computer as velocity

commands in a convenient coordinate frame (e.g., the commands can be

referenced with respect to the manipulator base, the end effector, or a

frame within a grasped object). Typically, each controller dof

corresponds to one spatial dof. as with direct rate.

Advantages: Allows a choice of coordinate frame. Relieves the operator

burden of coordinating joint activation. Can use linear and non-linear

gains. A small controller motion can cover large workspace accurately.

The accuracy of manipulator positioning does not depend on joint

resolution.

Disadvantages: End-effector location must be obtained visually from the

remote images, or via mental integration. Requires a moderate/high

degree of computation. Generally not compatible with force feedback.

15




Direct Unilateral Position Control

Under this control technique, the controller output is relayed directly to

the manipulator servo, where the signal is interpreted as the desired joint

motion. The controller dof typically correspond, on a one-to-one basis,

to the manipulator dof.

Advantages The controller input corresponds to the desired position of

the actuator. Simple implementation.

Disadvantages: Requires high resolution position sensors on both

controller and slave for electro-mechanical systems. Spatial

correspondence dependent on controller and manipulator configuration.

No force feedback. operator inputs can exceed the maximum velocity of

the arm. End-effector control frame cannot be specified. Limited use of

scaling.




Resolved Unilateral Position Control

Under this control scheme, the controller output is interpreted by a

computer as the desired spatial position and/or orientation of a

convenient coordinate frame attached to the manipulator (e.g., the end

effector or the pay-load). The computer converts the measured

controller signals into the equivalent Cartesian spatial movements of the

operator's hand, transforms the movement to the coordinate frame at the

slave control point.

Advantages: Choice of control coordinate frame. Spatial

correspondence can be achieved regardless of controller design. Motion

scaling can be incorporated.

Disadvantages. Needs a moderate/high degree of computation. Since the

controller configuration is not required to be the same as the arm

configuration, configuration feedback may not be available. Requires

high resolution position sensors on both controller and slave.




Direct Bilateral Position Control

Under this control scheme, the controller output is relayed directly to

the manipulator servo, where the signal is interpreted as a desired joint

motion. Simultaneously, the arm's actual joint positions sent directly to

the hand-controller servo, where it is interpreted as the required joint

position. This bilateral control results in force reflection in the hand-

controller and force generation in the slave arm, when the controller and

manipulator are in different positions.

Advantages. Controller input corresponds to the desired position of the

actuators. Simple implementation. Force feedback.

Disadvantages. Requires high-resolution position sensors on both

controller and slave for electro-mechanical systems. Spatial

configuration dependent on controller and manipulator configuration.

Increased controller complexity over unilateral position control. End-

effector control frame cannot be specified. Limited use of scaling.

16




Resolved Bilateral Position Control

The computer converts controller joint signals to an equivalent Cartesian spatial

movement of the operator's hand, transfers the movement to the control-point

coordinate frame of the remote manipulator, and solves for the joint commands

necessary to position the arm accordingly. Simultaneously, the computer

transforms the position and force encountered by the remote end-effector into

hand-controller coordinates, and determines the commands necessary to

position the hand-controller accordingly. Resolved bilateral control can also be

achieved by measuring the forces exerted by the slave directly and then

transforming those forces into feedback signals to the controller.

Advantages. Choice of control coordinate frame. Spatial correspondence can be

achieved regardless of controller design. Motion and force scaling can be easily

incorporated.

Disadvantages. High degree of computation necessary. Since controller

configuration is not required to be the same as the arm configuration,

configuration feedback may not be available. Requires high-resolution position

sensors on both controller and slave.




Filtering

Is a process in which extraneous motion that is

superimposed upon the control signal by the operator is

detected and subsequently deleted. Filtering can be very

useful when a miniature replica is being used.

Advantages. Removes unwanted control signals. Smooth

operator inputs.

Disadvantages. May remove desirable control signals. Can

introduce phase error. Moderate to high degree of

computation depending on filtering scheme.




Scaling

Scaling is a control aid in which the geometric gain between the

controller and the manipulator can be varied. A gain greater than one

allows a controller to perform gross motion over a workspace which is

larger than the control space. Conversely, a gain less than one allows the

same controller to perform precision motion with greater accuracy than

achievable with the unaided human hand.

Advantages. Single controller can perform both gross and precision

movements in limited control volume.

Disadvantages. Probability of operator error increased at high gains.

Extraneous input during high gain requires filter. Resolution of slave

must be at least that of controller resolution times the lowest gain.

Direct position control can only use scaling over limited regions without

loss of spatial correspondence.

17




Controller Re-referencing

Controller re-referencing is a control strategy in which the

operator can reference the control device with respect to the

control coordinates. One form of this technique maintains the

control device and its movements within an optimum volume, to

ensure that the operator can assume a comfortable and stable

configuration for his arm.

Advantages. Operator can work in physically and mentally

convenient coordinates.

Disadvantages. Discontinuity in control during changes. Operator

may loose spatial correspondence. Operator may experience

conceptual difficulty in switching between different coordinate

systems.




Control Coordinates Re-referencing

Control Coordinates re-referencing is a control strategy in which

the operator can change the control coordinate location. For

example, this technique is being used in the shuttle system to

allow changes between pay-load, end-effector, and orbiter-

located control coordinates.

Advantages. Operator can work in mentally-convenient control

coordinates. Can simplify tasks by working in natural

coordinates.

Disadvantages. Can only be used with resolved control

techniques. Moderate computational requirements. Provision

must be made for unique specification of desired control frame.




Motion Constraints

Motion constraints place artificial constraints on the manipulator

to either improve control or protect the system. Motion

constraints can be based on a model of the environment, directly-

sensed data, or both. Force accommodation is an example in

which control is improved through adaptive motion constraint

based on the forces and torques sensed at the end-effector.

Advantages. Improved control. Overall system protection.

Partially relives operator concern for system protection.

Simplifies operator input.

Disadvantages. Can require high degree of computation. Can

require a priori knowledge of the environment.

18




Compensation Techniques

Compensation techniques are a group of control strategies in which the

dynamic effects of the controller, manipulator, or task are removed or

compensated for to prevent burdening the operator and to improve

control. For example, a force/torque sensor could be mounted on the

controller handle and the measured operator force inputs could be used

to compensate for controller inertia and friction effects. Another

example of compensation is a control system which tracks the motions

of a moving task and superimposes that motion on the control signals,

effectively freezing the end effector in task coordinates.

Advantages. Unwanted effects can be removed from the system.

Disadvantages. Can require high degree of computation. Undesired

effect must be understood well enough to be compensated. Possible

danger of compensating important data.




Bilateral Position Control




Resolved Position Control

19



The FRHC Model X



Summary

• Introduction






• Conclusions



Verification of Teleop Systems

20




• Compute average force

• Compute maximum force magnitude

• Compute number of errors







21



Conclusions

• We have defined the problem of teleoperationand presented some of its relevant applications.

• A brief history of teleoperation development at NASA has been presented.

• The main design parameters of joysticks and relevant control structure have been discussed

• As an example of complex joystick, the JPL-Stanford FRHC has been presented.

• Some of the control strategies for teleoperation have been reviewed.

• An example of a verification and comparison of teleoperation control algorithms has been given.

Date post:	16-Apr-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

bertnoro03 fiorini 1 - PRISMA Lab · Paolo Fiorini Department of Infor mat i c s Uni ver s i t y of...

Documents