Sensing and Control within a Robotic End Effector

(1)

University of Southampton Research Repository

ePrints Soton

copyright owners. A copy can be downloaded for personal non-commercial

research or study, without prior permission or charge. This thesis cannot be

reproduced or quoted extensively from without first obtaining permission in writing

from the copyright holder/s. The content must not be changed in any way or sold

commercially in any format or medium without the formal permission of the

copyright holders.

When referring to this work, full bibliographic details including the author, title,

awarding institution and date of the thesis must be given e.g.

AUTHOR (year of submission) "Full thesis title", University of Southampton, name

of the University School or Department, PhD Thesis, pagination

(2)

UNIVERSITY OF SOUTHAMPTON

Faculty of Engineering and Applied Science

Department of Electrical Engineering

Sensing and Control within a Robotic End Effector

By

Venketeshwar Nath Dubey

A Thesis Submitted for the Degree of

DOCTOR OF PHILOSOPHY

(3)

UNIVERSITY OF SOUTHAMPTON

ABSTRACT

FACULTY OF ENGINEERING AND APPLIED SCIENCE

ELECTRICAL ENGINEERING

Doctor of Philosophy

SENSING AND CONTROL WITHIN A ROBOTIC END EFFECTOR

by Venketeshwar Nath Dubey

This research programme investigates aspects of end effector design and control, to carry out grasping operations in a range of unstructured environments.

A conceptual three fingered end effector design has been developed. The articulated finger is operated by a novel mechanism which provides all the finger motions. Detailed force and kinematic analyses have been carried out which establish mechanical integrity of the system and help size the various finger components. A vectorial method of link representation has been used to derive finger kinematics. This representation has been used for position control in the controller. A numerical technique based on the Newton-Raphson method has been derived to undertake the finger's inverse kinematics in real-time. To validate the theoretical operation of the finger drive, a mechanism has been built with the necessary electronic interface, and programmed for position control.

A photoelasticity based sensor has been developed which is capable of detecting applied force as well as slip and is largely immune to external disturbances. The sensor has a small size allowing it to be easily incorporated into a robotic finger. Mechanics of slip has been investigated to develop a theoretical model of the slip sensor. This allows modelling of various material and geometrical parameters involved in its design.

In order to control the end effector, grasping strategies have been planned and a controller structure defined. The top level of the controller uses the kinematic relation to move the finger to a goal position. When fingers make contact with an object, the controller switches over to an inner fiizzy logic algorithm. The rule base of the fiizzy logic ensures that a stable grasp has been acquired with minimum fingertip force. The implementation of the fuzzy logic has been validated on an experimental test-rig. It has been found that the controller applies different minimum fingertip force to objects of different mass and it responds very quickly to the external disturbances by applying extra force to the object. The fingertip force comes back to its previous level as soon as the disturbance vanishes. The important feature exhibited by the controller is that it forms optimal grasp of objects without knowing their mass and frictional properties. This offers a very useful capability to an end effector controller operating in unstructured environments.

(4)

(5)

IV

Acknowledgements

I would like to take this opportunity to acknowledge my sincere thanks to my

supervisors Dr. Richard M. Crowder and Dr. Paul H. Chappell for their help, support

and encouragement throughout the research. Their responsiveness and generous patience

will always be remembered.

I am deeply indebted to Dr. Richard L. Stoll for his effort to help me secure the research

award. Professor Adrian G. Bailey has been very kind to enquire about the progress of

the work from time to time. Thanks are due to Mr. David R. Whatley of the Central

Design Service for providing the manufacturing drawings of the finger mechanism.

Thanks to Dr. Suleiman M. Abu Sharkh for allowing me to test his software. I also

thank Dr. Paul M. Sharkey of Department of Cybernetics, University of Reading for his

comments on the structure of the thesis.

Many people have been instrumental in the research. Special thanks to Andrew Barton

for his help during programme debugging sessions. I also thank Hassan Pajooman for

offering practical tips on electronic circuit design. Gary Wills provided a matured

company and Colin Light's initiative to organise weekly research meetings are greatly

appreciated. Lindsey Whitmore, her laughter often reminded me that a life existed out of

my oflBce. Trevor Williams has been very generous to allow me to use his colour printer.

Martin Browne of Engineering Materials has extended an unconditional support to use

the friction testing machine and Peter Wheeler of Mechanical Engineering has always

been of immense help in lending mechanical tools.

Many thanks to Mrs. Mary Campbell for carefully listening to my problems and to the

other members of the department whom I have been in touch. Many people have made

my stay at Southampton a memorable one, Michael, Mohamed, Victor, Neamat and last

but not least Sabine. Finally, I am most grateflil to my father for his moral support and

endless patience, and thankful to my brothers and their families for encouragement.

This research has been supported by the financial grant from the Faculty of Engineering

and Applied Science, University of Southampton, Department of Electrical Engineering,

University of Southampton and the Overseas Research Student (ORS) award offered by

(9)

Chapter 1: Introduction

Chapter 1

Introduction

1.0 Background

Efforts are being made world wide to remove human operators fi-om hazardous

environments [TchemeshofF 1997]. Such environments typically include, space

exploration, underwater mining, and operations in nuclear and chemical industries. The

objective is to reduce or remove the risk to operator caused by the health and safety

hazards. In order to achieve this, robotic and automated systems need to be highly

flexible. However, the present level of technology does not provide a ready-made

solution for an autonomous system under such situations. The available autonomous

systems can cater only to the needs which are well defined and are known in advance.

For unstructured environments found in space, underwater and within nuclear

installation, where most of the objects to be handled are unknown and uncertain, a

robotic system has to make decisions based on its sensory information. The

non-deterministic nature of the environment, not only places a stringent requirement for any

robotic system developed, but it also demands a high level of intelligence to be

incorporated in such systems. The design of such systems will be an ultimate challenge to

human ingenuity.

The robotic system required to operate in such environment should be versatile enough

to handle unknown objects and situations. In other words, the gripper of the robotic

system should be able to grasp objects of different shapes and mass more like the human

hand than a conventional robotic end effector. This calls in for the design of a versatile

end effector with tactile sensing capabilities and capable of operating intelligently. In

order to develop a robotic system the three areas, namely the end effector design, tactile

sensing and control need to be considered in a greater depth. Once such a system is

developed, it can find application in the environment which poses a serious threat to

(10)

Chapter 1: Introduction

1.1 Previous Work on End Effectors

Research on multifingered end effectors has initially been inspired by the work on

prosthetic hands for the disabled and later the need to replace humans in hazardous

environment further promoted it. While currently most of the works in this area are

directed towards the fine manipulation of the object [Mason and Salisbury 1985],

[Fearing 1986] and [Akella et al. 1991], object handling with stable grasp is still of

considerable importance from practical application point of view. Crossley and Umholtz

[1977] have analysed that in the majority of instances a hand is used only to grasp

objects in an office or in a laboratory.

Many prosthetic hands have been reported [Kato 1982] which have the potential to be

used as practical end effectors. Notable among prosthetic hands are the hands from

Waseda and Tokyo Denki University. However, most of them do not use sensors and

have very low load carrying capacities. Out of the many end effectors which have been

designed and built, very few are reported to be used for practical applications. Three

commercially available end effectors are the Stanford/JPL, the Utah/MIT and the

Belgrade/use hands [Guo et al. 1992].

The design of the Stanford/JPL hand [Mason and Salisbury 1985] was motivated by the

anthropomorphic attention and was developed for use as a research tool in control and

design of articulated hands. The hand, as shown in Figure 1.1, has three fingers, each

with three degrees of fi^eedom. The hand is operated by twelve DC motors and twelve

tension cables.

(11)

Chapter 1; Introduction

Each finger mechanism has three joints, two parallel axis joints to provide curling action

and a third proximal joint, perpendicular to the other axes provides lateral motion. An

extra motor is used to effect extension of curled fingers thus, each finger needed four

actuators and four sets of tension cables. A solid state strain gauge provides information

on the finger tip force. The control of the Stanford/JPL hand is however, complex

because of the large number of motors and actuation system coupling. Also the whole

actuation system, comprising twelve motors, forms a large and heavy drive unit.

The Utah/MIT anthropomorphic hand [Jacobsen et al. 1986] has three fingers and an

opposing thumb as shown in Figure 1.2, each finger with four degrees of freedom. The

hand is powered by pneumatic cylinders through 32 independent polymetric tendons.

Each finger has three parallel axis joints to provide curling action and the fourth

perpendicular axis joint to provide lateral motion.

Figure 1.2 The Utah/MIT hand

The hand uses touch sensors based on the optical properties of the bifiingent material

with optical fibre links. The version IV of Utah/MIT hand [McCammon and Jacobsen

1990] uses of Hall effect sensors for position sensing as well as sensing the cable tension

for force measurement. A capacitive tactile sensing array has also been incorporated in

the hand. The large number of actuators requires a complex control. Curling and lateral

motion interfere and are not truly independent. This design is mainly intended to be used

(12)

Chapter 1: Introduction

The

Belgrade/use

hand [Bekey et al.

1990]

is a truly anthropomorphic end effector for robot manipulators with five fingers as shown in Figure 1.3. This operates on four

motors, two for fingers (one for each finger pair) and two for the thumb (one for lateral

and the other for curl/bend motion). Each finger has three parallel axis joints with a self

adaptability feature and with one degree of freedom to provide curling action with no

lateral motion.

Figure 1.3 The Belgrade/USC hand

The design uses a rocker arm which provides the self adaptability feature, when one

finger pad contacts the object surface, the other fingers continue to close until the

pressure on all the finger pads are approximately equal. The hand combines the motion

of finger segments to adapt to the object shape during grasping. DC servomotors located

within the wrist actuate the fingers through reduction gears directly with mechanical

linkages. Conductive plastic potentiometers are used as absolute position sensors for the

composite finger. Force sensing resistors (FSR) are used in finger tips and on the palm.

The high level grasp controller is knowledge-based, selecting a pre-shape on the basis of

visual information on the target and a stored library of relationships between grasp

modes and geometric primitives [Bekey et al. 1989].

Besides the above experimental robotic hands, research has been carried out at various

research laboratories to develop application-specific multifingered end effectors for

(13)

Chapter 1: Introduction

built for the NASA Skylab project, demonstrated a simple, reliable and versatile end

effector [Skinner 1975]. It has three fingers (Figure 1.4) each consists of three links and

the end effector operates on four motors.

Figure 1.4 The three-fingered end effector for NASA's Skylab project

The double-dwell turning mechanism rotates all fingers in different prehensile mode and

a separate motor provides curling motion to each finger with a cable-over-pulley system.

Various prehension modes can be obtained by just passing one control signal to each

motor. However, the operation restricts only specific prehension modes and the fingers

are not independent of each other, they cannot take any intermediate prehensile state. An

industrial version of this manipulator operates on two motors. One motor with a suitable

gear mechanism rotates two fingers in equal and opposite direction and the other motor

with a cross four bar chain mechanism bends all fingers. Each finger in this case is a

single mechanical link.

Crossley and Umholtz [1977] presented a detailed listing of the hand actions and

manipulation functions while considering a design for a three-fingered hand. They

identified that most of the grasp and manipulation can be realised with two fingers and a

thumb as an anthropomorphic model of the hand. The bending mechanism used in the

design of the finger is a turnbuckle mechanism which can produce a pinch force at the

fingertip. The gripping surfaces of the finger are cushioned with silicon rubber to adapt

(14)

Chapter 1; Introduction

to pick up thin surfaces from a table. The assembly was controlled by switches and

buttons to test validate the design feasibility.

Van Der Loos [1978] proposed a robust design for a multiple prehension three fingered

gripper. He analysed the various prehensile patterns and concluded that the mechanical

equivalent of this requires at least three fingers. In most cases of industrial grasping,

fingers have a symmetrical movement which can be achieved by a single actuator. But for

the sake of generality in grasping irregular and hollow objects, fingers need to be

operated with selective independence. He proposed the Cardan gearing assemblies for

two orientable fingers and an independent four bar linkage for the third finger. This

mechanism keeps the longitudinal axes of all three fingers parallel to each other and close

in a straight line perpendicular to the contact surface of the object. This design has

simple straight fingers and does not consider any curling or bending action of the finger.

Okada [1979] developed a three fingered hand with 11 degrees of freedom. The skeleton

sketch of the hand is shown in Figure 1.5, Each finger has four degrees of freedom and

the thumb has three degrees of freedom operated by cable and motors located in a box,

which is separate from the fingers. Fingers have a hollow and cylindrical shape which

makes them flexible and compact since the cable hoses and signal lines for the sensors

pass through the finger tubes.

• 3 3 l s : V

Figure 1.5 The Okada hand

The hand has primarily been developed as a test tool to perform assembly operations

which are based on hybrid posifion/torque control techniques. The hand has

(15)

Chapter 1; Introduction

opposing grip as position control and other as force control. Due to the cable/motor

system significant fiictional and elastic effects are produced in the system.

Some end effector designs have considered different types of actuation for controlling

robotic fingers. Nakano et al. [1984] have used shape memory alloy to control a three

fingered robot hand. Caldwell et al. [1995] have considered pneumatic muscle to control

robotic fingers. Robinson and Davies [1997] have developed an end effector for

underwater applications based on hydraulic systems. These end effectors have inherent

advantage of exhibiting high compliance, however, precise control of such systems are

difficult. Hence such actuation systems have not been considered in this research.

The mechanical design of the articulated hand developed at University of Bologna has

three fingers [Bonivento et al. 1988]. Two fingers have three segments each, with overall

six degrees of freedom plus one shared degree of freedom associated to the fingers for

the lateral motion. The other finger opposing the above fingers has two segments with

three degrees of freedom. Fingers are actuated by cables from DC motors and reduction

gearing. The hand uses potentiometer at the proximal joint as position sensor and strain

gauge as normal force sensor on each segment of fingers. The hand is mainly used for

grasping based on hybrid position and force control.

Design, control and construction of a three fingered, nine degrees of freedom robot hand

with built-in multicomponent force sensor have been described [Brussel et al. 1989]. The

hand has DC motors with planetary reduction gears imbedded into finger joints. This

introduces inertial force in the finger while it moves, and the whole system appears to be

bulky, however, it is shown to be capable of making fine manipulation of fragile object

like egg and performing peg insertion operations. The controller has a hierarchical

structure with three different levels of finger control, hand control and task control. The

hand incorporates a miniature incremental position encoder to the axis of each motor and

every finger is equipped with a three dimensional force sensor using strain gauges. The

fingertips have a rubber ball for fnctional point contacts. All connectors from motors,

(16)

Chapter 1; Introduction

The Karlsruhe hand [Doll and Schneebeli 1988] has three identical fingers each with

three degrees of fi-eedom. Each finger has two sections mounted on a revolute base

which has an axis perpendicular to the two distal joints axes. In order to avoid the

problem of friction and cable elongation, joints are driven directly by DC motors with

harmonic drives, in addition it uses timing belt for the two distal joints. This however,

introduces coupling between the distal joints. The earlier version used potentiometer at

joints for position sensing and strain gauges at the inner tubular section of the distal

segment for force measuring. However, the later version which is primarily meant for

manipulation, uses strain gauges at various locations for measuring pressure force,

torsion and bending moments. Besides this, it uses silicon rubber pad with air chambers

at the fingertip for fijrther tactile information [Magnussen and Doersam 1995]. The

earlier version used a hybrid position and force control scheme whereas the newer

version used a multilevel control system based on neuro-fiozzy algorithm on a parallel

computer.

A sophisticated five fingered whole arm manipulator (WAM) has been developed at the

University of Southampton for use within nuclear glovebox environment [Crowder

1991]. The end effector, shown in Figure 1.6 operates on three brushless DC motors and

the harmonic drive gear box placed inside the hand.

Figure 1.6 The Whole Arm Manipulator

Though it does not provide any lateral motion to the fingers, it is capable of forming

(17)

Chapter 1: Introduction

end effector has mechanically adaptive fingers with a mechanism which allows the

middle, ring and little finger to curl simultaneously whereas the index finger can curl

independently. The thumb is a single link unit which can open and close. A number of

optical reflecting touch sensors have been used and the control strategy is hierarchical

with different control levels. The whole end effector module has a rigid structure which

provides a practical solution for rugged and reliable use. However, this device does not

provide any manipulative feature where the control during finger extension operation is

difficult.

A gripper designed to operate in a hazardous environment has been developed at Delft

University of Technology [Jongkind 1993], has number of distinct features different from

other available grippers. The gripper has three fingers each with three segments and

three degrees of freedom as shown in Figure 1.7. Each finger can rotate about its base

axis at the proximal joint to form opposing mode as well as inside-out mode for hollow

or annular objects. The palm of the gripper has two degrees of freedom. The whole unit

uses 14 geared motors with pumps (remotely placed) and the hydraulic actuators located

inside the finger.

Figure 1.7 An end effector for hazardous environments

Information on position as well as applied force is obtained by using linear-variable

differential transformer (LVDT), resolvers are used for measuring angles. Position-force

(18)

Chapter 1: Introduction 10

National Taiwan University has recently developed a compact five fingered

anthropomorphic hand with 17 actuators to achieve dextrous manipulation [Lin and

Huang 1996], The thumb and the first finger has four degrees of fi-eedom whereas the

rest fingers have only three degrees of freedom. The actuators are the micro-motors

located inside each finger section with two sets of high ratio gear trains. Potentiometers

are installed at each joint as position sensor and 18 FSRs are used at various location of

hand for tactile sensing. The overall weight of the hand is reported to be around 1.6 kg

with a rated weight of object to be grasped is 1 kg. In order to manage large number of

actuators and sensors, adaptive fiizzy control has been used. The assembly of large

number of components in a compact design means the design may not be robust and

cannot be commissioned for an industrial use.

The review of various end effectors suggests that an end effector with three orientable

anthropomorphic fingers which can be controlled independently can meet most of the

grasping and manipulative function requirements. The frictional and elastic effects in end

effectors can be minimised if the tendon driven actuation is avoided. For control, weight

and costs considerations a minimum number of motor should be used. An end effector

incorporating above feature has been designed and reported [Crowder 1990], and this

design is the starting point for the present research.

1.2 The Target End Effector

In order to design and develop an end effector which is to operate in a target

environment the system specification needs to be identified clearly. Van Der Ham et al.

[1993] have conducted a survey among the users in nuclear industries to obtain the

fixture gripper needs in the nuclear environments. The survey covered the reliability

requirement of the system, task to be performed, object to be handled, accuracy,

environmental requirements, required grasps and mechanical and sensorial requirements

of the gripper. This survey indicates that the specification of a gripper, to operate in a

nuclear environment, is different from a gripper for general use. For example, the nuclear

environment typically has a level of radiation present which puts constraint on the

material selection of the gripper as well as sensors. A high reliability of the end effector

(19)

Chapter 1; Introduction 11

to operate in a moderately high temperature and needs to be capable of absorbing shocks

and vibration.

The original development of the current end effector is based on the design study carried

out by Crowder and Whatley [1987] for an in-reactor manipulator. Specifications of the

end effector were developed for the Central Electricity Generating Board (now Nuclear

Electric pic., UK), as a part of the Magnox reactor repair program, and hence is very

user specific. The design was developed to produce an adaptive end effector capable of

tackling most general type of objects during debris retrieval. Such an environment inside

the reactor is unstructured by its very nature where objects of different shapes, sizes and

mass are available. The retrieval operation does not require any manipulative function

nor any tool holding facility. Only versatile grasping capability of the end effector is of

paramount importance in such environments. The key specifications of the end effector

with these requirements are presented below which servers as a guide to the present

design.

• The end effector is to have three fully articulated fingers typically arranged

symmetrically (120° intervals). The fingers should be capable of handling unusually

shaped objects.

• The end effector should be capable of gripping a variety of objects for example, M6

to M30 nuts, 6 to 50 mm diameter bars up to 150 mm long picking them from a flat

surface, 50 mm cube, 100 mm diameter, 6 mm thick disc from a flat surface. A table

tennis ball is to be used as a test-object for force control.

• The maximum gripping force should be 100 N for a 100 mm diameter disc.

• The maximum overall diameter of the end effector is to be 120 mm, should weigh

less than 5 kg and should be as compact as possible.

• A closure time of 2 to 5 seconds needs to be achieved.

• Each finger should be capable of independent velocity, force and torque control.

• A range of tactile sensors needs to be incorporated in the end effector.

• Absolute position indicators are to be used.

• The unit is to be manufactured from stainless or low carbon steel and the structure

(20)

Chapter 1: Introduction 12

• The normal operating temperature is to be 100° C.

• In the case of power failure the fingers should be capable of being back driven.

While many of the above requirements will be met during the component selection and

construction phase, emphasis in this research has been given to establish a satisfactory

operation of each system component including the end effector design, tactile sensors

and the controller to help validate the concept design. Though the research has been

centred on these guidelines, the specification of maximum gripping force, the closure

time and the shock-absorbing capacity are particularly considered for component design

and for the selection of the drive unit.

1.3 Research Objectives

The aim of the research is to develop the technology of an end effector system including

tactile sensors and a controller to carry out grasping operation in an unstructured

environment in real time. The unstructured environment, in this context, means that

objects of different shape and mass (as in case of Magnox reactor debris) are available

whose properties such as mass, friction and rigidity are not known and the end effector

has to form an optimal grasp in each case. The approach taken in this research is to

identify a realistic design for practical implementations.

The research mainly focuses on three different aspects, the detailed analysis of an existing

three fingered end effector design, defining and developing a controller for the end

effector to operate in real time, and the assessment of the existing tactile sensors for use

in this context leading to the development of some special purpose tactile sensors. These

three areas, though very broad in themselves, have been pursued together as they are

inter-related for the specific application and can be better investigated if considered as a

whole. The specific key objectives of the complete research program are to analyse an

existing design of a three fingered end effector to verify the system parameters for

example, to check whether the component dimension of the end effector supports the

design or it requires some modifications. The kinematics required for controlling the end

effector and the capacity of the motor and brakes required to operate the end effector

needs to be evaluated. An interface, required for the end effector with suitable software

(21)

Chapter I: Introduction 13

In the sensing part of the work, a survey on the existing tactile sensors has to be

conducted to assess the suitability of these sensors in the robotic context. If the existing

sensors do not meet the specific requirements, development of some special purpose

tactile sensors needs to be investigated. Further to this, the experimentation with the

developed sensor, calibration and sensor integration into the end effector has to be well

addressed.

Since the controller is meant to operate in an unstructured environment in real-time, a

new controller structure needs to be defined to meet these requirements. In particular,

the controller should be able to meet the requirements of the end effector as well as it

should be able to process the sensor information in real time. In order to validate these

control ideas, experiments need to be conducted with the developed controller and its

viability need to be compared with other existing controllers.

Finally, a complete end effector model has to be developed incorporating the sensor and

the controller for the equilibrium and stability of the grasped object. This should further

consider the dynamic aspects of the grasp for applying minimum fingertip force to the

object and respond quickly to the external disturbances. The model should also be able to

minimise the system unbalance in position, force and moment due to uneven grasping

arising out in an unstructured environment and with the available structure of the end

effector fingers. The complete research programme is shown Figure 1.8.

1.4 Introduction to the Research

The Department of Electrical Engineering, University of Southampton has a long history

of research in the area of multifmgered robotic end effectors. Many end effectors have

been developed for a variety of needs and applications. Notable among them are, the

Whole Arm Manipulator developed for the UK Ministry of Defence and the in-reactor

end effector designed for the Central Electricity Generating Board, Both these end

effectors have similar structure of finger linkages however, in the whole arm

manipulator, the two inputs to a finger are provided through an equalising bar, which

offers an adaptable feature to the finger bringing in the curling or the bending action,

depending on the resistance experienced by the input linkages on the bar. Due to the

(22)

Chapter 1: Introduction 14

the control during finger extension has also proved to be difficult. The fingertip position

however, can be precisely controlled in the case of the in-reactor end effector where the

input to the two finger linkages can be provided independently or simultaneously.

Research objectives

End effector Tactile sensors

system design

force & kinematic analysis

Control

control strategies

controller structure

study on existing sensors sensor development

Modelling and simulation of the end effector

Figure 1.8 Aspects of the research

The two end effectors, mentioned above, have been designed by Central Design Service

(CDS), University of Southampton, which has a national reputation for innovative

system design in engineering. The design of such articulated fingers however, has not

been analysed thoroughly for the system kinematics and the force set up in each finger

components under various conditions of finger loading. Previously, an undergraduate

project had been undertaken in association with the CDS in this direction. The basic

(23)

Chapter 1; Introduction 15

dimensions. The present work has been developed from the above design concept to

carry out a thorough analysis of the in-reactor end effector to develop a complete model

of the system as a whole.

1.5 Preview

The following seven chapters in the thesis detail the work which has been carried out

during whole course of the research.

The design details of the three fingered end effector and the forward kinematics of the

articulated finger have been covered in Chapter 2. The chapter presents design features

of the end effector and a detailed kinematic and force derivations which are necessary to

relate the inputs to the finger with the output fingertip location and the corresponding

fingertip force. Following the design, kinematic and force derivations, a thorough

analysis has been carried out for various conditions of the finger loading and thus sizing

the various finger components and the motor drive. An interface has been developed for

the recently built finger drive mechanism to establish its satisfactory operation and to

conduct some trial runs. The chapter includes the details of the interface circuit and the

controlling software to operate the finger drive unit.

The inverse kinematics of the articulated finger has been derived in Chapter 3 which

relates the input displacements required by the finger to reach a target location. The

structure of the finger linkages has been geometrically analysed to identify the valid

linkage configurations for design and control purposes. The accuracy evaluation of the

implementation has been carried out and the flexibility of the finger motion has been

discussed with its design and control implications.

A comprehensive review of the existing tactile sensors is presented in Chapter 4. A

design of optical fibre tactile mesh has been proposed and the experimental studies on

force sensing resistors (FSR) have been included in order to assess their suitability for

use in the robotic end effectors.

In Chapter 5, design, construction and testing of a photoelasticity based sensor have been

(24)

Chapter 1; Introduction 16

investigations on the mechanism of slip and the design improvements of photoelasticity

based slip sensors to provide continuous signal during slip. Later, a theoretical model of

the sensor has been developed to analyse the experimental results from the sensor.

As the end effector is meant to operate in an unstructured environment for grasping

object of different shape and mass in real time, the aspect of grasping strategies, sensor

requirements and the controller structure have been dealt in Chapter 6. The controller

structure is based on the concepts of fuzzy logic to form a stable grasp of object. The

methodology and the experimental results of this implementation have been discussed.

The controller exhibits a novel feature of grasping objects of different mass with a

different minimum force without knowing the mass and the frictional property of the

grasped object.

In Chapter 7, a complete three dimensional model of the end effector has been developed

for the equilibrium of the grasped object minimising the unbalance in position, force and

the moment. The simulation results of this implementation have been discussed for the

various cases of grasping. This algorithm together with the fuzzy logic allows the

equilibrium, stability and the dynamics of the grasped object to be fully controlled, which

is capable of forming an optimal grasp with minimum fingertip force in each case.

The overall research achievements are discussed in Chapter 8. The conclusions drawn

from each chapter have been outlined with references to the relevant sections. The thesis

(25)

Chapter 2: End Effector: Design and Analysis 17

Chapter 2

End Effector : Design and Analysis

2.0 Introduction

An end eflfector is the generic term for a device at the end of a robot arm, which can

pick, manipulate, transfer or process work pieces, and includes spray guns, welding

torches, special purpose tooling and grippers. The end effector forms the mechanical

interface between a robot and the environment. Various types of end effectors for object

gripping are available with functionality ranging from a simple two jaw device to a

dextrous robot 'hand'. Repetitive work in an industrial environment can be performed

with a dedicated end effector capable of handling only one type of the work piece.

However, when shape, size and mass of the work pieces to be handled vary widely, a

cost effective implementation would be to use an advanced multifingered end effector

with a versatile grasping capability. Skinner [1986] has shown that for an automatic

assembly line, cycle time and the pay back period of a multiple prehension hand which is

capable of changing its gripping pattern by simply changing the finger bending directions

is lower than any other forms of gripper changing mechanisms e.g., quick-change fingers,

quick-change hands or the flip-over hands. Further, if the robotic system is to operate in

an unknown environment, it is apparent that the end effector should have attributes

similar to the human hand in functionality and sensing.

The complexity involved in duplicating the maximum handling, manipulating and sensing

capabilities of the human hand often prohibit a realistic end effector design to be

developed. In order to incorporate various manipulative functions, the end effector

needs to have large number of degrees of freedom which requires several independent

actuators. This approach makes the system bulky and complex to control with a

questionable reliability. If large number of sensors are used, interpreting the sensor

information may not be performed in real-time which may be warranted in some

applications. Thus to realise an end effector for practical use with good reliability, only

the minimum essential requirements need to be considered. The present work is based

(26)

Chapter 2; End Effector ; Design and Analysis 18

In this chapter the design of an end effector and the kinematic analysis of its articulated

finger are presented. The objective of this analysis is to identify the mechanical forces set

up in the finger components as a result of the applied gripping force, and to investigate

the kinematics of the system with a view to define a position control algorithm for a

controller. A solution has been developed to identify the force experienced by various

links and joints when the finger interacts with the environment.

2.1 The End Effector Design

Figure 2.1 shows the details of the finger drive mechanism developed for use in the three

fingered end effector. The end effector is a self contained unit designed to be attached

directly to a robot. It consists of three equally spaced fingers. Each module is identical

and are mounted within the body of the end effector. The finger module, as shown in the

sectional front view of the figure, has a vertical frame and is pivoted at its central axis by

two stub-shafts containing a motor, which is rigidly fixed to the main body of the

gripper. The articulated finger, shown in Figure 2.2, is attached on top of the mechanism

and is actuated by the two leadscrew shafts within the mechanism. The two leadscrews

are driven by timing belts from the two outputs of the differential gearbox. The gearbox

is connected by another timing belt to the motor output shaft such that the housing of the

differential is a driven unit. The differential is supported on bearings to allow the two

pinion outputs to be transferred to the leadscrews (from the top and the bottom of the

differential unit). Electromagnetic brakes are employed at each shafts (one on each

leadscrew shaft, and one at the bottom frame shaft) to direct the motor output to one or

more of these driven units. Thus the mechanism can split an input motion into three

different components to drive the finger in different axes.

The articulated finger has three sections hinged at joints C and F. Sections I and 3 are

connected by control link 4 at joints E and G. Control link 3 connects the section 2 with

the bell crank at joints D and I. Control link 1 connects to a rectangular nut and the bell

crank at joints M and K. Control link 2 connects the section 1 at joint B with another

rectangular nut at joint L. Joints A and J are hinged and are fixed to the frame. All the

other joints are hinged and are floating with respect to the frame. Input to the finger is

provided from the nuts on the two vertical leadscrews as displacements d l and d2 (in the

(27)

Chapter 2: E n d Efifector: D e s i g n and Analysis 19

1 7 4 , 0

view showing t h r e e f i n g e r m o d u l e

4 . 0

leadscrew n u t s l e a d s c r e w s

b r a k e I b r a k e 2

b r a k e 3 8&5

is=

a

14 5.0

d i f f e r e n l i a l u n i t

30.0-f r o n t view (in s e c t i o n ) side view

(28)

Chapter 2: End Effector : Design and Analysis 20

control link 2

rGctarxgxilar nuts

section 3

section 2 control link 4

H (X.Y)

confrof krw: 7

s e c t i o n 1

control link 3

df cK

bell crank

lower travel limit

Figure 2.2. The articulated finger

The finger mechanism provides three degrees of fi'eedom to a finger. Two of these are

for curling (where displacement of control link 1, rotates two upper finger sections about

joint C) and bending (where displacement of control link 2 bends all the three finger

sections about joint ^4) as shown in Figure 2.3. The third motion is for rotating the whole

module about its central axis relative to the end effector body.

(29)

Chapter 2; End Effector : Design and Analysis 21

A complete finger module is shown in Figure 2.4. The structure of the finger offers

independent curl motion, while the bend motion is only partially independent which

brings in slight curling effect in the finger. Since each component of motion is driven by a

single motor in combination with three electromagnetic brakes through the differential

gearing mechanism, the three components of motions can be driven individually or in

combination.

F-IJ

front vie* (in seclion) aide view

(30)

Chapter 2: End Effector: Design and Analysis 22

Thus the finger can be operated in an adaptable mode when two leadscrew brakes are

OFF and in a precise control mode when the leadscrew brakes are operated individually.

If the three brakes are designated as brake 1, brake! and brakeS for curl, bend, and rotate

motions of the finger respectively then the resulting finger motion can be shown in Table

2.1.

Table 2.1 The finger motions

No Brake 1 Brake 2 Brake 3 Resulting finger motion Typical applications

1 OFF OFF OFF curl, bend and rotate moving to home position

2 OFF ON ON curl final wrap-around grasp

3 ON OFF ON bend initial finger approach

4 ON ON OFF rotate finger orientation

5 OFF OFF ON curl and bend fast approach and grasping

6 ON OFF OFF bend and rotate grasping configurations

7 OFF ON OFF curl and rotate manipulation

8 ON ON ON none finger lock

Ideally, all the above finger motions are possible, however the finger motions relating to

the rotation of the finger module in combination with the other motions (No 1, 6, 7) is

difficult to achieve due to the high inertia and friction of the whole mechanism in which

case, the curl and the bend motion always precede the rotate motion.

The geometry and the dimensions of the finger components are shown in Appendix A,

and are those used for the following analysis. The shape of the fingertip has been chosen

to simplify the force derivation. However, in practice the finger surface is curved for

smoothly transferring the contact force to the object, from one finger orientation to the

other and keeping the fingertip-normal perpendicular at the contact surface. The two

fingertip angles tipl and tip2 have been shown for construction details but in the analysis

only the tipl angle has been used to calculate the fingertip normal. The depth of the

finger sections 1 and 2 to be used in the analysis {MiPj and M2P2 in Figures 2.5 and 2.6)

have not been specified in the dimensioned figure in the appendix. However, for the

(31)

25-Chapter 2: End Effector; Design and Analysis 23

20% more than the dimension u (dimension u is shown in Appendix A). Further the

normal force to these sections are considered to be acting at their mid-point

(length-wise). The finger origin specifies the origin of the finger about which all references are

made and this designates the zero displacement for the upper travel limit of the nuts. The

lower travel limit is the maximum possible displacement of the nuts. With the specified

size of the nut, the effective leadscrew displacement possible is limited to 0.0339 m. In

the following analysis the skeleton structure of the finger sections have been shown,

however some of the referenced symbols are not shown. Readers are referred to the

finger dimension and angle nomenclature diagrams in Appendix A, for the symbols which

are not available in the specific figure. This has been done for clarity and to avoid the

repetition of the finger parts.

2.2 Force and Kinematic Analysis of the Articulated Finger

As the object is grasped by the robotic gripper, the contact between the fingers and the

object results in interaction forces and moments. This subjects the finger components to

different levels of loading, especially the finger joints. Since the finger is driven by

actuators, the corresponding input torque transmitted to a fingertip needs to be

calculated. In other words, for a known interaction of finger with the environment, the

loading of the various finger components needs to be identified for the mechanical design

as well as the force control implementation. Also in order to achieve a specific grasping

task through the movement of fingers, the study of the motion of the finger geometry is

important to validate the design feasibility and for the implementation of position control

algorithms.

The homogenous matrix transformation method has been extensively used to derive the

kinematic relations for articulated links [Asada and Slotine 1986]. The static force

analysis for a closed loop kinematic chain has been carried out [Mason and Salisbury

1985] with multiple fingers grasping an object. A three finger design has been described

using an heuristic combination of position and force control [Okada 1986]. However in

the present design, a vector method has been used for static force and kinematic

derivations since the finger forms a planer mechanism which can be described

(32)

Chapter 2: End EfiFector; Design and Analysis 24

2.2.1 Analysis Overview

The following assumptions have been made for the analysis:

• fingertip has a point contact

• friction and viscous forces are considered to be negligible in the mechanism

• no compliance or backlash are present in the system

• system is quasi-static

• mechanism is planer (since it executes a planer trajectory)

• analysis is based on the specified dimension of the finger component (Appendix A)

These assumptions allow a linear model of the system to be developed. A Cartesian

coordinate system is used throughout in the analysis, with the coordinate frames fixed to

various finger joints. The joint axis positions are described as Cartesian vectors such as jc

and y components of that vector. For example a vector AC has two component vectors

ACx^Tid ACy in x a.ndy directions respectively. The angles are treated as vectors to avoid

misinterpret the vector directions and are taken positive in the anticlockwise sense with

respect to the x-axis. In the geometrical section of the analysis, the vector addition and

subtraction rules are followed throughout to relate each component of the finger.

Extensive use of the cosine rule has been made to calculate the included angle between

two vectors and of the inverse trigonometric functions, to extract the vector angles from

the Cartesian vectors. Due to the high degree of mobility of some of the links, care has

been taken to ensure that the computed angles fall in the right quadrant. Apart from this,

the main technique used consists of the application of the laws of equilibrium and matrix

method for solving sets of simultaneous linear equations.

2.2.2 Equilibrium Equations

As shown in Figure 2.3, each finger can move to various locations in space and can adapt

to grasp an object. The tip of each finger alone could touch an object and forces are

produced in the finger components. In order to study the loading on each part of a

finger, a force analysis has been made with 1 N forces applied to the tips and to the mid

point of the other sections both individually and collectively. For calculating the force on

each component, equilibrium equations are derived from the free body diagram of each

(33)

Chapter 2: End Effector : Design and Analysis 25

convenient point to give three equations for each link. The four links, three main finger

sections and the bell crank, give non-trivial equations, providing a total of 12 equations

in 12 unknown forces. Thus the equations are soluble.

The initial analysis has been carried out by considering a normalised force of 1 N acting

normally to the fingertip. All the equilibrium equations are linearly related to the force (as

will be seen later) and can be scaled with respect to 1 N. For a range of applied forces at

the fingertip, the actual force existing on various finger components is obtained by

multiplying the computed force by a factor by which the fingertip force has been

exceeded in comparison to the IN force. However, the analysis also considers the

different forces applied to various sections at different possible inclinations to identify the

critical conditions of loading and of the finger components.

2.2.2.1 Equilibrium of finger section J

The free body diagram of the finger section 1 is shown in Figure 2.5. All capital letters

pointing to an arrow are force vectors inclined at certain vector angles (Greek letters).

Note that the name of the joints and the link parameters are the same as the original

nomenclature in Figure 2.2.

A normal force WJ is applied at the middle point Pi of the section (mid point of A and E)

to account for the force arising out of the finger interaction with an object in a wrap

around grasp. The location of the point Pi is based on,

' 2

= 1.25w (25% more than the distance between joints A and B)

ZAP^ = tan - 1

1 /

AP, +

Resolving forces horizontally,

[/ + 7Vcosr = g + / ( c o s g + PFIsina 2.1

Resolving vertically,

PFl cosa + f + # sin r = F + 7^ sin g 2.2

(34)

Chapter 2: End Efifector; Design and Analysis 26

( 1 8 0 - a )

( T —(J + 90)

( 9 0 —£ + a )

\R

Figure 2.5 Forces on finger section 1

clockwise moments = aP cos a + aQsina + PFl sin(90 - ZAP^)AP^

anticlockwise moments = UNCOS{T - cj + 90) + (a + f)Rcos(s -90-a)

Thus for equilibrium,

aPcosa + aQsma+W\sin(90 - ZAP^)AP^ =

uNcos{r -(7 + 90) + (a + / ) ^ c o s ( g - 90 - a ) 2.3

2.2.2.2 Equilibrium of finger section 2

Figure 2.6 shows the forces acting on the section 2 of the finger, with the same

assumption a normal force W2 is acting at the middle point P2 of the section 2 at a

location,

CM, = -' 2

M2P2 = 1-2m (20% more than the distance between joints A and B)

ZCP. = tan'

V CMj /

CP^ =4CMI

Q + F cosO= T+ Wliv[\p

P-vS- F smO-^r W2cos/?

2.4

(35)

Chapter 2: End Effector: Design and Analysis 27

Taking moment of forces about C,

CP2 W2 sin(90 - Z-CPj ) + bTsin = bScosp + cF cos(90 -y + 6) 2.6

(90-ICP2 }

( 9 0 — y + Q )

Figure 2.6 Forces on finger section 2

2.2.2.3 Equilibrium of finger section 3

The two angles S and p, as shown in Figure 2.7, are the bend and contact component of

the fingertip which changes when the fingertip bends about joint F and when the contact

location on the fingertip changes respectively.

Equilibrium of section 3 with a normal force W can be derived as,

ZLHHl - K- tip\ + 5 y/ = ZHHl - nil

W c o s x j / = - T - Rcoss 2.7

sin (y = -6" - T^sin g 2.8

Taking moment of forces about F,

(36)

Chapter 2; End Effector; Design and Analysis 28

TT— tvp 1+6

Figure 2.7 Forces on section 3 (fingertip)

2.2.2.4 Equilibrium of the bell crank

From the forces acting on the bell crank (Figure 2.8), the equilibrium equation can be

written as.

F cos9 - Z + Xcosco 2.10

F smO+ Y = Xsinco 2.11

Taking moment of forces about J,

JF cos(A -d) = kXcos(co -A- ZJJK) 2,12

(u-X-flJK) /

(37)

Chapter 2: End Effector : Design and Analysis 29

2.2.2.5 Matrix representation of equilibrium equations

Equations (2.1-2.12) can be represented in the matrix form such that,

[A][B]=[C]

where [A] is the matrix of geometrical coefficients which is.

2.13

0 -COSi 0 1 COSf 0 0 -1 0 0 0 0

0 sinz 1 0 -sinf 0 0 0 -1 0 0 0

0 _A3.2 a c o s a a s i n a •^3,5 0 0 0 0 0 0 0

COS 6 0 0 1 0 0 -1 0 0 0 0 0

-sind 0 1 0 0 1 0 0 0 0 0 0

Afi,] 0 0 0 0 _{bcosfi -bsin/3} 0 0 0 0 0

0 0 0 0 -COSf 0 -1 0 0 0 0 0

0 0 0 0 -sinf -1 0 0 0 0 0 0

0 0 0 0 dcosis-S) 0 0 0 0 0 0 0

COS0 0 0 0 0 0 0 0 0 -COSO) 0 -1

s i n 0 0 0 0 0 0 0 0 0 -sin<a 1 0

A) 2,1 0 0 0 0 0 0 0 0 _AI2,10 0 0

Ax2~ums(i-a+90), Axs=-{a+f}cos(s-90-a), ki^\=cco5(90-y+ Ai2,i=;'cosfl-^, Ai2,io= -kcos(a}-X-ZIJK)

[B] is the column vector of unknown forces

[B]^=: ^ N P Q R S T U V X Y Z ^

[C ] is the column vector of known forces with geometrical coefficients

[ C ] ^ = |~-W7sina -fVJcosa C3.1 IV2sinP IV2cosjff Cgj (fcosy/ fysiny/ Cg,, 0 0 ^

C3,i= -lVJsin(PO-j^Pj)APi, C6.i-fV2sin(P0-^P2)CP2, C9,i=IVgsin(^-p)

From equation 2.13, [B] can be obtained as,

[ B ] = [ A r [ C ] 2.14

However, as the values contained in [A] and [C] are largely functions of the vector

angles between the linkages, the relationships must be found between these angles and

the two positional input variables dJ and d2.

2.2.3 Vector Angle Derivations

(38)

Chapter 2; End Efifector; Design and Analysis 30

Referring to the linkages skeleton diagram (Figure 2.9), angles r, a, and a are the

functions of the leadscrew displacement d2 which can be geometrically related as.

C ^ E

Figure 2.9 Definition of angles r, crand a

LAy = LOy + OAy = d2 S — T

LA^ =q-p (for definition of s, r, q, p see figure in Appendix A)

ZLA - {LAy jLAx)

And, \LA\= ^LAl +LAl

Now applying the cosine rule to the triangle LAB,

T = COS-'({\LA\'+h'-u')/{2\LA\h)) + ZLA ( j = 7t + ZLA- cos"'((|L^|'W -h')l{l\LA\v^ a - a- ZBAC

(ZBAC is a known finger parameter, refer Appendix A)

2.15

2.16

2.17

2.2.3.2 Derivation of angles od and A

From Figure 2.10, it can be seen that angle X and co are functions of displacement dl.

Since, MJ^ = q

(39)

Chapter 2; End Effector ; Design and Analysis 31

(for q and r refer Appendix A)

ZMJ = tan -I MI. \MJ\= ^MJl+MJ,

d1

Figure 2.10 Definition of angles (O and X

(o = cos-^((lMjf+/^ - k^) / (2|M/|/)) + ZMJ

X = 7r/2 + ZMJ - ZIJK - cos~^((|M/l^+yt^ - p-)! (2\MJ\k))

2.18

2.19

2.2.3.3 Derivation of angle 6

Angle 9, as shown in Figure 2.11, can be found by applying the cosine rule to the triangle

ICD.

since, AJ^ = p

AJy =-s

(for p and s refer Appendix A)

JI^ = - j s i n A

Jly = 7 cos/I

(40)

Chapter 2; End Effector : Design and Analysis 32

AC^ = flcosa

ACy = a sin a

IC = AC - AJ^

Figure 2.11 Definition of angle 9

>\IC\= ^ICl+lC

ZIC = tan"

e = c o s - ' ( ( v ' + | / C f V ) / (2v|/C|)) + ZIC 2.20

2.2.3.4 Derivation of angles rand p

Y is the angle of vector CD (Figure 2 . 1 1 ) .

since, ID^ = v cos 6

IDy = vsin^

CD=ID-1C.

C D , = Z D , - / C ,

(41)

Chapter 2; End Effector; Design and Analysis 33

fi=y-ZI)CF 111

{ZDCF is a known finger parameter, refer Appendix A) 2.2.3.5 Derivation of angles & d and p

From Figure 2.11,

= v 4 C . + C D ,

+CD^

AE^ -{a+ / ) c o s a

AEy = (a + / ) sin a

CF^ - b cos p

>|£F|= ^EF'^ +EF^

ZEF = tan -1 EF,\ y

Now referring to Figure 2.12,

g = cos-'((|EF|"+g"-(/")/(2e|EF|)) + Z E F 2.23

since, S is the angle between vector FG and the vertical,

AG^ = AE^ +gcosg

AGy = AEy + e sin £•

F G , = j G , - / U ?

ZFG = tan

\FGJ

5-ZFG-nll 2.24

(42)

Chapter 2: End Effector : Design and Analysis 34

{Z.GFH is a known finger parameter, refer Appendix A)

Figure 2.12 Definition of angles s, S and p

2.2.4 Kinematic Relations

The vector angles derived in the preceding sections are used to describe kinematic

relations for Cartesian position of the fingertip (AH^, AH^ as shown in Figure 2.13.

y

a """fa A

Figure 2.13 Definition of the fingertip position

AH^ = + CF^ + FH^ = a cos a + cosy9 + gcos p

AHy = ACy + CFy+FHy = flslH u + ^ sl H p+gsin p

2 26

(43)

Chapter 2: End EflFector : Design and Analysis 35

where

a = ;r + tan - 1 d2 •¥ S — T

Sensing and Control within a Robotic End Effector

University of Southampton Research Repository

reproduced or quoted extensively from without first obtaining permission in writing

awarding institution and date of the thesis must be given e.g.

Venketeshwar Nath Dubey

UNIVERSITY OF SOUTHAMPTON

CONTENTS

CHAPTER 1

CHAPTER 3

CHAPTER 5

CHAPTER 7

Conclusions 203

Acknowledgements

Chapter 1: Introduction

Chapter 1

1.0 Background

Chapter 1: Introduction

Chapter 1; Introduction

Chapter 1: Introduction

Belgrade/use

Chapter 1: Introduction

Chapter 1; Introduction

Chapter 1; Introduction

Chapter 1; Introduction

Chapter 1: Introduction

Chapter 1: Introduction 10

1.2 The Target End Effector

Chapter 1; Introduction 11

Chapter 1: Introduction 12

1.3 Research Objectives

Chapter I: Introduction 13

1.4 Introduction to the Research

Chapter 1: Introduction 14

Chapter 1; Introduction 15

Chapter 1; Introduction 16

Chapter 2: End Effector: Design and Analysis 17

Chapter 2

2.0 Introduction

Chapter 2; End Effector ; Design and Analysis 18

2.1 The End Effector Design

Chapter 2: E n d Efifector: D e s i g n and Analysis 19

front vie* (in seclion) aide view

25-Chapter 2: End Effector; Design and Analysis 23

2.2 Force and Kinematic Analysis of the Articulated Finger

Chapter 2: End EfiFector; Design and Analysis 24

2.2.2 Equilibrium Equations

Chapter 2: End Effector : Design and Analysis 25

CP^ =4CMI

Chapter 2; End Effector; Design and Analysis 28

Chapter 2: End Effector : Design and Analysis 29

COS0 0 0 0 0 0 0 0 0 -COSO) 0 -1

s i n 0 0 0 0 0 0 0 0 0 -sin<a 1 0

2.2.3 Vector Angle Derivations

Chapter 2; End Effector ; Design and Analysis 31

Figure 2.11 Definition of angle 9

g = cos-'((|EF|"+g"-(/")/(2e|EF|)) + Z E F 2.23

2.2.4 Kinematic Relations

Chapter 2: End EflFector : Design and Analysis 35