CPS: Cadd Pro Solutions


"Think differently to achieve great success,
This is the time to awake,
This is the time to develop a new era..."

Robotics

Introduction

The word “robot” was introduced by the Czech playright Karel ˇ Capek in his 1920 play Rossum’s
Universal Robots. The word “robota” in Czech means simply “work.” In spite of such practical beginnings,
science fiction writers and early Hollywood movies have given us a romantic notion of robots.
Thus, in the 1960s robots held out great promises for miraculously revolutionizing industry overnight.
In fact, many of the more far-fetched expectations from robots have failed to materialize. For instance,
in underwater assembly and oil mining, teleoperated robots are very difficult to manipulate and have
largely been replaced or augmented by “smart” quick-fit couplings that simplify the assembly task.
However, through good design practices and painstaking attention to detail, engineers have succeeded
in applying robotic systems to a wide variety of industrial and manufacturing situations where the
environment is structured or predictable. Today, through developments in computers and artificial intelligence
techniques and often motivated by the space program, we are on the verge of another breakthrough
in robotics that will afford some levels of autonomy in unstructured environments.
On a practical level, robots are distinguished from other electromechanical motion equipment by their
dexterous manipulation capability in that robots can work, position, and move tools and other objects
with far greater dexterity than other machines found in the factory. Process robot systems are functional
components with grippers, end effectors, sensors, and process equipment organized to perform a controlled
sequence of tasks to execute a process — they require sophisticated control systems.
The first successful commercial implementation of process robotics was in the U.S. automobile
industry. The word “automation” was coined in the 1940s at Ford Motor Company, as a contraction of
“automatic motivation.” By 1985 thousands of spot welding, machine loading, and material handling
applications were working reliably. It is no longer possible to mass produce automobiles while meeting
currently accepted quality and cost levels without using robots. By the beginning of 1995 there were
over 25,000 robots in use in the U.S. automobile industry. More are applied to spot welding than any
other process. For all applications and industries, the world’s stock of robots is expected to exceed
1,000,000 units by 1999.
The single most important factor in robot technology development to date has been the use of
microprocessor-based control. By 1975 microprocessor controllers for robots made programming and
executing coordinated motion of complex multiple degrees-of-freedom (DOF) robots practical and
reliable. The robot industry experienced rapid growth and humans were replaced in several manufacturing
processes requiring tool and/or workpiece manipulation. As a result the immediate and cumulative
dangers of exposure of workers to manipulation-related hazards once accepted as necessary costs have
been removed.
A distinguishing feature of robotics is its multidisciplinary nature — to successfully design robotic
systems one must have a grasp of electrical, mechanical, industrial, and computer engineering, as well
as economics and business practices. The purpose of this chapter is to provide a background in all these
areas so that design for robotic applications may be confronted from a position of insight and confidence.
The material covered here falls into two broad areas: function and analysis of the single robot, and
design and analysis of robot-based systems and workcells.
Section 14.2 presents the available configurations of commercial robot manipulators, with Section
14.3 providing a follow-on in mathematical terms of basic robot geometric issues. The next four sections
provide particulars in end-effectors and tooling, sensors and actuators, robot programming languages,
and dynamics and real-time control. Section 14.8 deals with planning and intelligent control. The next
three sections cover the design of robotic systems for manufacturing and material handling. Specifically,
Section 14.9 covers workcell layout and part feeding, Section 14.10 covers product design and economic
analysis, and Section 14.11 deals with manufacturing and industrial processes. The final section deals
with some special classes of robots including mobile robots, lightweight flexible arms, and the versatile
parallel-link arms including the Stewart platform.

Design of Robotic Systems

For manufacturing in which the manufacturing facility is concerned with similar volumes of production
and a wider range of parts, the assembly line/mass production method is often not cost effective. It is
often desirable to group equipment units together into workcells that can, in composite, perform an
entire family of related operations on the product. The work-in-progress enters the workcell, remains
while several functions are performed, and then leaves the workcell.
The individual equipment units that are used in the workcell (for both processing and materials
handling) can consist of combinations of manual, semiautomatic, and fully automated equipment.
However, in this section, the term “workcell” refers to a grouping of the robot and its peripheral equipment
to assemble any of a large variety of products with little or no human intervention, driven by electronically
designed data. An assembly robot is a comparatively simple mechanism whose function is to position
parts and tools in the space of its work volume accurately. It is a comparatively low-cost machine of
high precision of positioning and great reliability. Its simplicity, however, excludes the possibility of
human-type actions like form recognition and its prehensile tools are very far from having the number
of degrees of liberty a human hand has. If we concede that an assembly robot can by no means compete
with a human being in a complex task, we also have to acknowledge that an assembly robot is capable
of executing monotonous tasks with consistently high precision, thereby increasing the quality of the
product. It can also keep up a fast production line indefinitely. Recognizing this difference between a
human and a robot is essential in the design of a robotic system.
The remainder of the section is organized as follows. A set of design considerations for designing an
assembly robot workcell is first presented. Layouts for a typical robotic workcell are then discussed.
Experience so far has shown that in most instances, it is a feeder that fails in the workcell, not the robot.
Feeding methods must be carefully considered when designing a workcell and are discussed at the end
of the section.


Workcell Design and Layout

Design Considerations


Assembly systems can be broadly classified as manual, fixed, and flexible systems in relation to the
complexity of the product to be assembled and to the production volume as shown in Figure 14.9.1.
Flexible robotic workcells are typically used for less complex products at low or medium production
volume, while for increasing product complexity, the cells designed for a single-purpose task can be
linked into assembly lines. Apart from product volume and complexity, the design of the robotic workcell
depends on several factors: namely, number of part types, end-of-arm tooling exchange, as well as
product design.
Number of Part Types. A typical workcell consists of a robot and its peripherals made up of partpresentation
mechanisms, feeders, conveyor, and end-of-arm tooling. For a small number of part types,
parts are presented to the robot by feeders or magazines. As these take up space, only a limited number
of different parts can be fed to one robot. In mechanical assembly normally a maximum of five to six
different parts can be presented in this way.
To extend the robot’s accessibility to a large number of parts, mechanized component feeding systems
can be mounted on data-driven carousel conveyors spaced around the robot, each with a fixed dispensing
point within reach of the robot gripper. The carousel can accommodate up to several hundred positions
onto which magazines, tapes, or other modular dispensing systems can be attached. With multiple
programmable carousels, the robot can access several thousand different parts. The application of the
mechanized carousel is useful when only a few of each part type from thousands of styles may be used.
Other alternatives are (1) kitting, in which all components to be assembled are kitted in a loosely

palletized waffle pack, and followed by more accurate location using standard machine vision; and (2)
the use of accurate totes for robot handling.
End-of-Arm Tooling Exchange. Many systems use different gripper exchange systems in order to cope
with different parts. Tool exchanges are often considered as “nonproductive” since they do not contribute
to assembly operations. The exchange is serially coupled to the assembly operations. This means that
the cycle time increases due to the extra time needed for pickup and drop-off for tool changes as well
as travel time between the assembly point and the end-of-arm tooling station. In order to reduce time
loss due to the gripper, exchange should be minimized and/or in parallel with other activities, and the
distance between pick-up point and assembly point should be very short. This problem could be avoided
if a fast-revolving gripper head is used provided that space, weight, and cost of the revolving head do
not pose a problem. Alternatively, the pallet carries batch-specific equipment such as grippers, fixtures,
and end-of-arm tooling and can be presented to the robot on a conveyer in a similar fashion as the parts.
Product Design. Product design for flexible automation cells includes the following criteria: task
operations based on flexible assembly cells for specific product families which must be able to assemble
the variants of these product families using programming, fast changeover from one product to another
within a flexible assembly cell, and reuse of standard elements for new assembly tasks
In addition to product complexity and volume, two other criteria should be considered in the construction
of flexible assembly cells. First, since only a few products are generally suitable for fully
automatic assembly, manual working processes are often essential with a large number of products.
Flexible assembly cells must be constructed so that manual work stations can be included following
ergonomic principles. Second, since the type-specific peripheral costs will increase in relation to the
number of individual parts in the product to be assembled, part-specific feeders must be minimized for
the economic use of flexible assembly cells.

 Workcell Layout

Workcell design and layout in a flexible automation system depend on the nature of the manufacturing
processes, the product design, and the material handling system as a whole. The manufacturing systems
are classified as electronic product assembly, subassembly of electrical and mechanical components, and
kitting cell for large-scale manufacturing.
Electronic Product Assembly. Flexible workcells are commonly used for the assembly of integrated
circuit boards (PCB), where a combination of interchangeable part-feeding mechanisms are used to
present parts to robots. Since a majority of the processes involved are carried out in the linear, vertical
plane, robots of SCARA or gantry construction are best suited for these assembly tasks. The workcell consists of a robot and its peripherals made up of part-presentation mechanisms, feeders, conveyor, and
end-of-arm tooling.
Figure 14.9.2 shows the organization of a typical workcell for assembly of a family of circuit boards
(Decelle, 1988), which is a part of the in-line component insertion, inspection, and repair assembly line.
Circuit boards to be assembled are secured on panels and flow through the workcell on a conveyor. Each
of the circuit boards is characterized by a bar-coded serial number that permits product tracking, data
collection, and testing through the assembly. Boards requiring assembly are positioned over an elevator
mechanism in the workspace of the robot. The mechanism lifts the board slightly and uses the tooling
holes on the panel to locate the circuit board. Two digital signals interface the conveyor to the workcell
— one signals the robot that the board is ready for assembly and the other signals the conveyor to index
the board to the next workcell. Components are fed to the robot by using feeders. The feeder singles
out components to a walking-beam mechanism that transfers parts through lead-cutting, lead-straightening,
and lead-verification operations and on to the lead-locating nest for robot pickup. The activities
of the robot and its peripherals in a workcell are coordinated by a host computer. The workcell is set
up and monitored through the host computer. Through the host computer, the operator provides the
workcell the code to be assembled, the components in the feeders, and the configuration of the feeders.

 Subassembly of Electrical and Mechanical Components. Unlike PCB assembly, packaging and
designs of small electrical and mechanical components are generally nonstandardized. Thus, the problem
of automated flexible assembly workcells lies in the presentation of parts and the degree of flexibility
of assembly of small components often involves both product design and layout considerations extensively.
Figure 14.9.3 shows a self-contained flexible workcell for assembly of small mechanical parts with
a circular indexing table. Modular part-feeding equipment such as vibrator feeder bowls and specialpurpose
trays are placed around the indexing table to feed and to orient small components to the robot.
Typical mechanical operations such as riveting, screwing, welding, inserting, pressing, and so on are
achieved through quick changeover end-of-arm tooling. The circular indexing table arrangement is
advantageous where end-of-arm tooling changes are necessary for handling different parts. It allows
changes of end-of-arm tooling to take place while other operations are continuing.
With complex products, assembly in a single flexible assembly cell is not always feasible. In this
case, a flexible assembly line can be designed to link self-contained independent workcells (Figure
14.9.3) so that they can be engaged or disengaged as required to allow adaptability in connection with
product model change. Alternatively, standard carriers or pallets can be used to present a large number
of different part-types to a robot. Each pallet carries a large number of identical parts, unoriented but

 

 with the right side up, and placed on a flat board. Standard machine vision was used to detect the
orientation of the parts.
Kitting Cell for Large-Scale Manufacturing. In the field of large-scale manufacturing such as automobile
manufacturing, engine assembly, and machining processes, where the setup time of specialized tools
for each task is excessive, the work is generally distributed into several cycle zones. As an example,
actual cutting time (production time) represents a value between 5 and 20% of average machine utilization
time that includes nonproductive time accountable by workpiece load/unload, too) change/setting, and
workpiece inspect.
To avoid a high level of wear and tear on tools due to constant conversion, the cycle zone is commonly
divided into individual operating cells which may be interconnected in series, parallel, or a combination
of series and parallel. A typical workcell (Figure 14.9.3) consists of a robot, a part-feeder, an end-ofarm
tooling section, and the manufacturing process. The parts are contained in a regularly spaced pallet,
which are transported by means of an automated guided vehicle (AGV) or a conveyor to the loading
tables and are fed to process by the robot. The most common approach in automated part presentation
for machine loading is the use of specially designed pallets for each part family to maintain sufficient
position accuracy for a completely preprogrammed robot picking.
In the case of assembly, purchased parts or parts to be processed are kitted onto one kit tray in a
single location. Kitting is the process of taking parts from bulk and placing them on a kit tray, which
is an organized group of parts. Concentrating the material delivery system and its control to one area is
the main benefit of the kitting cell. In addition to efficient use of floor space by eliminating duplicate
equipment at each assembly cell, the feeders and tooling are universal — the same equipment is being
used all the time for all parts, thus maximizing utilization while minimizing capital expense. The material
delivery equipment is eliminated at the assembly cycle times. Also, having all the parts for an assembly
on a carrier permits changes in the process route during machine downtime or blockages.
Figure 14.9.4 shows a layout of the kitting cell. An overhead gantry takes bins of parts and dumps
them in the appropriate feeders (indicated in Figure 14.9.4 as F1, … F7). The feeders fill the lanes with
an initial quantity and replenish them as parts are kitted. The parts come to rest in nests at the end of
the feeder lanes. Here the vision system verifies the correct part family, performs some quality checks,
and determines the position and orientation for the robot to pick the parts. Should the vision reject the
part, the nest will dump the part and a new part will be fed in for an inspection. Using a quick changeover
gripper, multiple parts are kitted onto a tray. Once all the parts are on the kit tray, the tray is indexed to
the inspection station for verification that all parts are placed. The robot takes the completed kit tray
and places it on the assembly conveyor to an idle station, ready to be picked up by an AGV.


Part-Feeding and Transfers

The term “part-feeding” refers here to feeding workpieces from pallets using a preprogrammed robot
for subsequent processes such as machining or assembly. The cost to feed parts to a robot for either 

machine loading or assembly in a flexible manufacturing system (FMS) has often been underestimated,
which may comprise as high as two thirds of the overall investment and is usually the source of a large
percentage of work stoppages and defects. A general review of existing mechanical part feeders can be
found in Lee (1991).
The basic kinds of part-feeding may be classified as follows: (1) mechanical feeders which are designed
to feed and to orient the parts-dedicated part-feeding apparatus, (2) dimensionally dedicated pallets
which are specially designed for each part family to maintain the position/orientation, and (3) machine
vision.

Mechanical Feeders

The commonly used mechanical feeders for robotic assembly are bowl feeders, vibratory feeders, and
programmable belt feeders. For large volume manufacturing, the employment of the dedicated mechanical
part-feeding apparatus may be justified. However, mechanical feeders consume a lot of room around
the workcell, often fail due to jamming, and, most significantly, generally require retooling when a
component is changed or tool wear is caused by jamming.
Vibratory Bowl Feeders. Vibratory bowl feeders (Boothroyd and Dewhurst, 1985) are most commonly
used as mechanical feeders for robotic assembly. The basic component of a bowl feeder consists of a vibratory bowl, a rotating disk, and an orienting track. Parts to be fed to the robot are separated into a
single line and oriented to move to feeding end. These feeders, in general, are not designed to be easily
converted to feed new part types. The cost of the bowl feeders can be broadly divided into two parts:
special purpose equipment cost and general purpose equipment cost. Typically, changeover would involve
replacing the bowl, orientation track, feed track, and escapement, which contribute to special-purpose
equipment cost. Only the vibratory drive unit could be reused. This general-purpose portion of the feeder
is approximately 30% of the feeder cost.
One way to lower the cost of the bowl feeder per part is to deliver different parts to a robot assembly
station using multiple layer vibratory bowl feeders. A multilayer bowl feeder has several bowls mounted
in stacked fashion, and in each bowl a different kind of part is stored. The design of multipart vibratory
feeders aims at reducing the cost of the vibratory feeders by sharing the general-purpose hardware cost
over several parts and by reducing the special-purpose tooling cost. Two basic forms are available: bowl
type and in-line type.
To change over this multipart vibratory feeder to other part types, the orienting tracks must be replaced.
An effective way to reduce wear is to separate the function of orienting from feeding. The function of
the multilayer vibratory feeders is to restrict feeding parts to a separation unit. Parts of several different
types are fed but not oriented from a vibratory feeder. In most cases, the workpieces must be held by a
mechanical pusher against a pair of orthogonal datum planes on a relatively flat surface with the “right
side up.” A machine vision system is then used to locate and/or to sort the orientation of the parts using
two-dimensional binary images, which is a great deal easier to store and to process.
Vibratory Belt Feeders. In vibratory belt feeders, parts are fed by a vibratory conveyer belt (Boothroyd
and Dewhurst, 1985). The principle of the vibratory belt feeder is to produce a vibratory motion on the
surface of the brushplate. The motion is obtained by pulling the brushplate sharply down and back and
then allowing it to spring up and forward. This action, repeated at high speeds (approximately 3600
times per minute at 60 Hz power supply), produces a definite vibrating movement on the brushplate
surface, permitting parts to be conveyed in a smooth and easily controlled manner.
The orienting systems used on these belt feeders may be a mechanical device, an optical sensor, or
a vision system. A machine vision system is often used to locate and/or to sort the orientation of the
parts using two-dimensional binary images. A line-scan camera is commonly used to create the silhouettes
of the workpieces, and in some cases the product designs can be reviewed to simplify the vision algorithm
and to reduce the system cost. Since a robot gripper can grasp parts from a queue on the feeder itself,
belt feeders do not require any special-purpose tooling for feed track or escapement and thus offer several
advantages over the vibratory bowl feeder for robotic assembly.

Dimensionally Specific Pallets

One of the most common approaches is the use of specially designed trays or totes for each part family
to maintain sufficient accuracy for a completely preprogrammed robot picking. A particular form of
these dimensionally precise feeders is known as tape-and-reel for feeding parts of relatively small sizes,
which can be placed on tapes of standard width. For some devices that are large, heavy, ceramic, or
have fragile leads, tapes are very expensive and impractical.
In general, the dimensionally specific pallets are well suited for large volume production where changes
of part types are not frequent. The operational cost of the design-specific pallets includes packaging
costs for transport, construction cost for pallet alignment, and engineering cost for new pallet designs.

Vision-Based Flexible Part-Feeding

For flexible manufacturing, where a large variety of product sizes and component types are encountered,
the part-feeding system must have the ability to adapt to a changing product design without costly
hardware redesign or time-consuming software reengineering. This need has been addressed as a general
industrial vision-based bin-picking problem by several authors.
In manufacturing automation applications, the processing speed of acquiring and analyzing an image
must be comparable to the speed of execution of the specific task. The attempt to duplicate human perception by obtaining a three-dimensional detailed image of the part often calls for time-consuming
computation and does not necessarily determine the location and orientation of a given part with the
accuracy required for successful part acquisition by the robot. Moderate location inaccuracies pose no
difficulty for human operators since they use vision, hand-eye coordination, and sense of touch to locate
and correctly load the part.
However, if the orientation of the parts can be characterized by the two-dimensional object silhouette,
retroreflective materials can be used as a background in generic part presentation (Lee and Li, 1991).
Most surfaces on objects exhibit a combination of diffuse and specular reflection. A point on an ideal
diffuse-reflecting surface appears equally bright from all viewing directions. Surfaces covered with papers
and matte paints may be considered as reasonable approximations. An ideal specular reflector is one
that reflects any incident light ray as a single ray in the same plane as the incident ray and the surface
normal. The basic principle of the retroreflective vision sensing is to structure the surface reflectance of
the pallet or the landmarks so that it is much brighter than objects generally characterized by diffuse or
specular surfaces. In practice, a number of nonpredictable factors such as measurement noise, the
uniformity of the surface reflectivity, and the uniformity of illumination, which occur on both the object
and the background, can be eliminated by a relatively simple technique. If part design can be modified,
brightly illuminated retroreflective landmarks can be intentionally created on objects for location tracking.
Low cost landmarks could be incorporated in design by using retroreflective liquid paints on existing
features. Alternatively, generic landmarks can be constructed by applying solid glass beads on the
reflected surface of standard fastening devices such as screw heads.

Robot Manufacturing Applications
John W. Priest and G. T. Stevens, Jr.

Product Design for Robot Automation

Identifying automation opportunities early in product design is important because product design requirements
to facilitate robotic manufacturing are often unique and must be integrated early in the product
design process. Some overall manufacturing problems for using robots and some design solutions to
resolve these problems are listed in Table 14.10.1.


Rossi highlighted the product designer’s role in robotics stating this problem (Rossi, 1985):

Often designs are made in such a fashion that one cannot access a certain area with a robot. Humans can get around
obstacles and operate within those designs easily, but robots cannot because they are not quite as flexible as human
beings. I think that this is the single most important item that has kept us from being further along than we are. What
happens is that users try to apply a robot to something that’s been designed without robotic assembly in mind. They
usually run into a problem. Either the robot cannot handle it at all or the users find that they have got to put a lot of
additional engineering design into a particular workcell, or perhaps into an end effector, in order to get around the
problem. All this does is add to the price tag, and cost is very much in consideration when one is trying to sell these
systems. A situation arises where robots are no longer attractive because of all the additional things that need to be done.

In summary, the product must be designed for the manufacturing process and the robot. For more
information, the reader should review Boothroyd (1994), Bralla (1986), Priest (1988), and Tanner (1994).
Table 14.10.2 shows some design rules for robotic assembly.

 Economic Analysis

Economic analyses for robotic applications are similar to those for any manufacturing equipment
purchase and usually use minimum annual revenue requirements, present value methods, or break-even
analyses. Since robots are a flexible method of automation, a unique aspect of robotics is manufacturing’s
ability to reuse the robot after its initial production run for other applications in later years. For many
companies, this subsequent use of the robot can be shown in the economic evaluation. Some other unique
benefits in robotic economic analysis that may be included are improved quality, higher precision, ability
to run longer shifts, and reduced floor space. Unfortunately, some unique disadvantages of robot analysis
are software integration complexity, inability to respond quickly to product design changes, and process
reliability.
In general, there are several situations where robots are more likely to make economic sense. These are
A. Sufficient volume to spread investment costs over many units
1. High volume
2. Stable product design
3. Multishift operations
B. Robot is used on more than one product
1. Limited number of different products on same production line
C. Part handling problems occur when performed manually
1. Parts that are very large, heavy, or bulky
2. Parts that are very fragile or easily damaged
3. Parts that are extremely small
D. Extremely difficult manufacturing process without using robot or automation
1. Many processes, especially in electronics, cannot be performed without robots or some type of
automation
E. Safety and health concerns of process
1. Safety and health costs can be significant
The type of data concerning the robot system that is required for an economic analysis is shown in
Table 14.10.3.

 Defining Terms

Accuracy. The degree to which the actual and commanded positions (of, e.g., a robot manipulator)
correspond for computed as opposed to taught positions.
Adaptive Control. A large class of control algorithms where the controller has its own internal
dynamics and so is capable of learning the unknown dynamics of the robot arm, thus improving
performance over time.
AML. A Manufacturing Language — a robot programming language.
APT. Automatic Programming of Tools — a robot programming language.
Cell Decomposition. An approach to path planning where the obstacles are modeled as polygons and
the free space is decomposed into cells such that a straight line path can be generated between any two
points in a cell.
Compliance. The inverse of “stiffness” — useful in end effectors tooling whenever a robot must
interact with rigid constraints in the environment.
Computed-Torque Control. An important and large class of robot arm controller algorithms that
relies on subtracting out some or most of the dynamical nonlinearities using feedforward compensation
terms including, e.g., gravity, friction, coriolis, and desired acceleration feedforward.
End Effector: Portion of robot (typically at end of chain of links) designed to contact world:
• Compound. A cluster of multiple end effectors and tooling mounted on the robot wrist.
Defining Terms
Accuracy. The degree to which the actual and commanded positions (of, e.g., a robot manipulator)
correspond for computed as opposed to taught positions.
Adaptive Control. A large class of control algorithms where the controller has its own internal
dynamics and so is capable of learning the unknown dynamics of the robot arm, thus improving
performance over time.
AML. A Manufacturing Language — a robot programming language.
APT. Automatic Programming of Tools — a robot programming language.
Cell Decomposition. An approach to path planning where the obstacles are modeled as polygons and
the free space is decomposed into cells such that a straight line path can be generated between any two
points in a cell.
Compliance. The inverse of “stiffness” — useful in end effectors tooling whenever a robot must
interact with rigid constraints in the environment.
Computed-Torque Control. An important and large class of robot arm controller algorithms that
relies on subtracting out some or most of the dynamical nonlinearities using feedforward compensation
terms including, e.g., gravity, friction, coriolis, and desired acceleration feedforward.
End Effector: Portion of robot (typically at end of chain of links) designed to contact world:
• Compound. A cluster of multiple end effectors and tooling mounted on the robot wrist.
• Active. An end effector with sensing and servo control of the grasp forces and/or finger motions.
• Prehensile. An end effector that holds parts between fingertips or encircled by fingers.
• Vacuum. A nonprehensile end effector that uses suction cups to hold parts.
• Dextrous. A hand with the ability to manipulate parts in the fingers and actively control grasp
forces.
Feedback Linearization. A modern approach to robot arm control that formalizes computed-torque
control mathematically, allowing formal proofs of stability and design of advanced algorithms using
Lyapunov and other techniques.
Flexible-Link Robot. Lightweight mechanical structures where vibration and flexibility of the links
must be taken into account in controller design. They possess favorable features including lower manufacturing
costs, higher motion speeds, better performance, and easier transportation and setup.
Force Control. A class of algorithms allowing control over the force applied by a robot arm, often
in a direction normal to a prescribed surface while the position trajectory is controlled in the plane of
the surface.
Forward Kinematics. Identification of task coordinates given configuration.
Fuzzy Logic Control. A multilevel logic controller, which is different from the conventional dual
(two-level) logic in which only two values (true and false) may be assigned to each state variable. Fuzzy
logic controllers have advantages in being robust to disturbances and not requiring an explicit mathematical
model for the design process. They consist of three parts: the fuzzifier, the rulebase, and the
defuzzifier.
Grasp Isotropy. A measure of how uniformly forces and motions can be controlled in different
directions.
IGES. International Graphics Exchange Specification — a data exchange standard.
Inverse Kinematics. Identification of possible configurations given task coordinates.
I/O Device. Input/output device — a port through which external information is connected to a
computer. I/O devices may be A/D, which converts analog signals to digital, D/A, which converts digital
signals to analog, or binary, which passes digital signals.
Joint Variables. Scalars specifying position of each joint — one for each degree of freedom.
Kitting. The process of taking parts from bulk and placing them on a kit tray, which is an organized
group of all parts required to assemble a single product or subassembly. An end effector with sensing and servo control of the grasp forces and/or finger motions.
• Prehensile. An end effector that holds parts between fingertips or encircled by fingers.
• Vacuum. A nonprehensile end effector that uses suction cups to hold parts.
• Dextrous. A hand with the ability to manipulate parts in the fingers and actively control grasp
forces.
Feedback Linearization. A modern approach to robot arm control that formalizes computed-torque
control mathematically, allowing formal proofs of stability and design of advanced algorithms using
Lyapunov and other techniques.
Flexible-Link Robot. Lightweight mechanical structures where vibration and flexibility of the links
must be taken into account in controller design. They possess favorable features including lower manufacturing
costs, higher motion speeds, better performance, and easier transportation and setup.
Force Control. A class of algorithms allowing control over the force applied by a robot arm, often
in a direction normal to a prescribed surface while the position trajectory is controlled in the plane of
the surface.
Forward Kinematics. Identification of task coordinates given configuration.
Fuzzy Logic Control. A multilevel logic controller, which is different from the conventional dual
(two-level) logic in which only two values (true and false) may be assigned to each state variable. Fuzzy
logic controllers have advantages in being robust to disturbances and not requiring an explicit mathematical
model for the design process. They consist of three parts: the fuzzifier, the rulebase, and the
defuzzifier.
Grasp Isotropy. A measure of how uniformly forces and motions can be controlled in different
directions.
IGES. International Graphics Exchange Specification — a data exchange standard.
Inverse Kinematics. Identification of possible configurations given task coordinates.
I/O Device. Input/output device — a port through which external information is connected to a
computer. I/O devices may be A/D, which converts analog signals to digital, D/A, which converts digital
signals to analog, or binary, which passes digital signals.
Joint Variables. Scalars specifying position of each joint — one for each degree of freedom.
Kitting. The process of taking parts from bulk and placing them on a kit tray, which is an organized
group of all parts required to assemble a single product or subassembly.

Learning Control. A class of control algorithms for repetitive motion applications (e.g., spray
painting) where information on the errors during one run is used to improve performance during the
next run.
Linearity in the Parameters. A property of the robot arm dynamics, important in controller design,
where the nonlinearities are linear in the unknown parameters such as unknown masses and friction
coefficients.
Manipulator Jacobian. Matrix relating joint velocities to task coordinate velocities - configuration
dependent.
Mechanical Part Feeders. Mechanical devices for feeding parts to a robot with a specified frequency
and orientation. They are classified as vibratory bowl feeders, vibratory belt feeders, and programmable
belt feeders.
Mobile Robot. A special type of manipulator which is not bolted to the floor but can move. Based
on different driving mechanisms, mobile robots can be further classified as wheeled mobile robots,
legged mobile robots, treaded mobile robots, underwater mobile robots, and aerial vehicles.
Path Planning. The process of finding a continuous path from an initial robot configuration to a goal
configuration without collision.
PD-Gravity Control. A special case of computed-torque control where there is a PD outer control
loop plus a gravity compensation inner control loop that makes the DC values of the tracking errors
equal to zero.
Pinch Grasp. A grasp in which a part is clamped between fingertips.
Pixel. Picture element — one point of an image matrix in image processing terminology.
Prismatic Joint. Sliding robot joint which produces relative translation of the connected links.
Redundant Manipulator. Manipulator for which the number of joint variables is greater than the
number of task coordinates.
Remote-Center Compliance (RCC). A compliant wrist or end effector designed so that task-related
forces and moments produce deflections with a one-to-one correspondence (i.e., without side effects).
This property simplifies programming of assembly and related tasks.
Revolute Joint. Rotary robot joint producing relative rotation of the connected links.
Robot Axis. A direction of travel or rotation usually associated with a degree of freedom of motion.
Robot Joint. A mechanism which connects the structural links of a robot manipulator together while
allowing relative motion.
Robot Link. The rigid structural elements of a robot manipulator that are joined to form and arm.
Robust Control. A large class of control algorithms where the controller is generally nondynamic,
but contains information on the maximum possible modeling uncertainties so that the tracking errors
are kept small, often at the expense of large control effort. The tracking performance does not improve
over time so the errors never go to zero.
SCARA. Selectively compliant assembly robot arm.
SET. (Specification for Exchange of Text) — a data exchange standard.
Singularity. Configuration for which the manipulator jacobian has less than full rank.
Skew Symmetry. A property of the dynamics of rigid-link robot arms, important in controller design,
stating that – 1/2Vm is skew symmetric, with M the inertia matrix and Vm the coriolis/centripetal
matrix. This is equivalent to stating that the internal forces do no work.
Stewart Platform Manipulator. A special type of parallel-link robot consisting of six identical linear
actuators in parallel, an upper platform, and a base. One end of each actuator connects to the base, and
the other to the upper platform with two- or three-degrees-of-freedom joints. This manipulator has a
greater force-to-weight ratio and finer positioning accuracy than any commercial serial-link robot.
Task Coordinates. Variables in a frame most suited to describing the task to be performed by
manipulator.
VDAFS. (Virtual Data Acquisition and File Specification) — a data exchange standard.

Visibility Graph. A road map approach to path planning where the obstacles are modeled as polygons.
The visibility graph has nodes given by the vertices of the polygons, the initial point, and the goal point.
The links are straight line segments connecting the nodes without intersecting any obstacles.
Voronoi Diagram. A road map approach to path planning where the obstacles are modeled as
polygons. The Voronoi diagram consists of line having an equal distance from adjacent obstacles; it is
composed of straight lines and parabolas.
Wrap Grasp. A grasp in which fingers envelope a part, to sustain greater loads.

References

Anderson, R.J. and Spong, M.W. 1989. Bilateral control of teleoperators with time delay. IEEE Trans.
Robotics Automation. 34(5):494–501.
Asfahl, C.R. 1992. Robotics and Manufacturing Automation. 2nd ed. John Wiley & Sons, New York.
Ballard, D.H. and Brown, C.M. 1982. Computer Vision. Prentice-Hall, Englewood Cliffs, NJ.
Book, W.J. 1984. Recursive Lagrangian dynamics of flexible manipulator arms. Int. J. Robotics Res.
3(3):87–101.
Boothroyd, G. and Dewhurst, P. 1985. Part presentation costs in robot assembly. Assembly Automation,
138–146.
Boothroyd, G., Dewhurst, P., and Knight, W. 1994. Product Design for Manufacture and Assembly.
Marcel Dekker, New York.
Bottema, O. and Roth, B. 1979. Theoretical Kinematics, North Holland, Amsterdam.
Bralla, J.G. (ed.). 1986. Handbook of Product Design for Manufacturing, McGraw-Hill, New York, 7-
75, 7-100.
Craig, J. 1985. Adaptive Control of Mechanical Manipulators. Addison-Wesley, Reading, MA.
Craig, J.J. 1989. Introduction to Robotics: Mechanics and Control. Addison-Wesley, Reading, MA.
Critchlow, A.J. 1985. Introduction to Robotics. Macmillan, New York.
Decelle, L.S. 1988. Design of a robotic workstation for component insertions. ATEJT Tech. J.
67(2):15–22.
Denavit, J. and Hartenberg, R.S. 1955. A kinematic notation for lower-pair mechanisms based on
matrices, J. Appl. Mech. 22:215–221,
Duffy, J. 1980. Analysis of Mechanisms and Robot Manipulators, John Wiley & Sons, New York.
Elfes, A. 1987. Sonar-based real-world mapping and navigation. IEEE J. Robotics Automation. RA-
3(3):249–265.
Fraden, J. 1993. AIP Handbook Of Modern Sensors, Physics, Design, and Applications. American
Institute of Physics, New York.
Fu, K.S., Gonzalez, R.C., and Lee, C.S.G. 1987. Robotics. McGraw-Hill, New York.
Fuller, J.L. 1991. Robotics: Introduction, Programming, and Projects. Macmillan, New York.
GMF Robotics Training and Documentation Department. 1985. Paint Processing: Concepts and Practices.
GMF Robotics Corporation, Troy, MI.
Groover, M.K., Weiss, M., Nagel, R.N., and Odrey, N.G. 1986. Industrial Robotics: Technology, Programming,
and Applications. McGraw-Hill, New York.
Gruver, W.A., Soroka, B.I., and Craig, J.J. 1984. Industrial robot programming languages: a comparative
evaluation. IEEE Trans. Syst. Man Cybernetics. SMC-14(4).
Hayati, S., Tso, K., and Lee, T. 1989. Dual arm coordination and control. Robotics. 5(4):333–344.
Hollerbach, J.M., Hunter, I.W., and Ballantyne, J. 1992. A comparative analysis of actuator technologies
for robotics. In Robotics Review 2, O. Khatib, J. Craig, and T. Lozano-Perez, (eds.). MIT Press,
Cambridge, MA, 299–342.
Hollis, R.L., Allan, A.P., and Salcudean, S. 1988. A six degree-of-freedom magnetically levitated variable
compliance fine motion wrist. In Robotics Research, the 4th Int. Symp., R. Bolles and B. Roth.,
(eds.). MIT Press, Cambridge, MA, 65–73.

Jacobsen, S., Wood, J., Knutti, D.F., and Biggers, K.B. 1984. The Utah/M.I.T. dextrous hand: work in
progress. Int. J. Robotics Res. 3(4):Winter.
Jamshidi, M., Lumia, R., Mullins, J., and Shahinpoor, M. Robotics and Manufacturing: Recent Trends
in Research, Education, and Applications, Vol. 4. ASME Press, New York.
Klafter, R.D., Chmielewski, T.A., and Negin, M. 1989. Robotic Engineering: An Integrated Approach.
Prentice-Hall, Englewood Cliffs, NJ.
Latombe, J.C. 1991. Robot Motion Planning. Kluwer Academic Publishers, Amsterdam.
Lee, K.-M. 1991. Flexible part-feeding system for machine loading and assembly. I. A state-of-the-art
survey. II. A cost-effective solution. Int. J. Prod. Economics. 25:141–168.
Lee, K.-M. and Blenis, R. 1994. Design concept and prototype development of a flexible integrated
vision system, J. Robotic Syst., 11(5):387–398.
Lee, K.-M. and Li, D. 1991. Retroreflective vision sensing for generic part presentation. J. Robotic
Syst. 8(1):55–73.
Leu, M.C. 1985. Robotics software systems. Rob. Comput. Integr. Manuf. 2(1):1–12.
Lewis, F.L., Abdallah, C.T., and Dawson, D.M. 1993. Control of Robot Manipulators. Macmillan, New
York.
Lewis, F.L., Liu, K., and Yesildirek, A. 1995. Neural net robot controller with guaranteed tracking
performance. IEEE Trans. Neural Networks. 6(3):703–715.
Liu, K., Fitzgerald, J.M., and Fewis, F.L. 1993. Kinematic analysis of a Stewart platform manipulator.
IEEE Trans. Ind. Electronics. 40(2):282–293.
Lozano-Perez, T. 1983. Robot programming. Proc. IEEE. 71(7):821–841.
McClamroch, N.H. and Wang, D. 1988. Feedback stabilization and tracking of constrained robots. IEEE
Trans. Automat. Control. 33:419–426.
Mujtaba, M.S. 1982. The AL robot programming language. Comput. Eng. (2):77–86.
Nichols, H.R. and Lee, M.H. 1989. A survey of robot tactile sensing technology. Int. J. Robotics Res.
8(3):3–30.
Nomura, H. and Middle, J.E. 1994. Sensors and Control Systems in Arc Welding. Chapman and Hall,
2-6 Boundary Row, London SE1 8HN, UK.
Okabe, A., Boots, B., and Sugihara, K. 1992. Spatial Tessellations, Concepts and Application of Voronoi
Diagrams. John Wiley & Sons, New York.
Pertin-Trocaz, J. 1989. Grasping: a state of the art. In The Robotics Review 1, O. Khatib, J. Craig, and
T. Lozano-Perez, Eds. MIT Press, Cambridge, MA, 71–98.
Priest, J.W. 1988. Engineering Design for Producibility and Reliability. Marcel Dekker, New York.
Rossi, M. 1985. Dialogues. Manuf. Eng. October: 41:24.
Sandler, B.Z. 1991. Robotics, Designing the Mechanisms for Automated Machinery. Prentice-Hall,
Englewood Cliffs, NJ.
Shimano, B.E., Geschke, C.C., and Spalding, C.H., III. 1984. Val-II: a new robot control system for
automatic manufacturing. Proc. Int. Conf. Robotics. March 13–15:278–292.
Siciliano, B. 1990. Kinematic control of redundant robot manipulators: a tutorial. J. Intelligent Robotic
Syst. 3(3):201–210.
SILMA Incorporated. 1992. SILMA CimStation Robotics Technical Overview. SILMA Incorporated,
1601 Saratoga-Sunnyvale Road, Cupertino, CA.
Slotine, J.-J. 1988. Putting physics in control: the example of robotics. Control Syst. Mag. 8 (December):
12–15.
Snyder, W.E. 1985. Industrial Robots: Computer Interfacing and Control. Prentice-Hall, Englewood
Cliffs, NJ.
Spong, M.W. and Vidyasagar, M. 1989. Robot Dynamics and Control. John Wiley & Sons, New York.
Stauffer, R.N. 1984. Robotic assembly. Robotics Today. October.
Stevens, G.T. 1994. The Economic Analysis of Capital Expenditures for Managers and Engineers. Ginn
Press, Needham Heights. MA.

Stewart, D. 1965. A platform with six degrees of freedom. Proc. Inst. Mech. Engr. (London)
180(15):371–386.
Tanner, W.R. 1994. Product design and production planning. In CRC Handbook for Robotics. CRC Press,
Boca Raton, FL, 537.
Taylor, R.H., Summers, P.D., and Meyers, J.M. 1982. AML: a manufacturing language. Int. J. Robotics
Res. (1):19–41.
Tsai, R.Y. and Lenz, R.K. 1989. A new technique for fully autonomous and efficient 3D robotics hand/eye
calibration. IEEE Trans. Robotics Automation. 5(3):345–358.
Tzou, H.S. and Fukuda, T. 1992. Precision Sensors, Actuators, and Systems. Kluwer Academic, Publishers,
Amsterdam.
Winston, P.H. 1984. Artificial Intelligence. Addison-Wesley, Reading, MA.
Wright, P.K. and Cutkosky, M.R. 1985. Design of grippers. In The Handbook of Industrial Robotics, S.
Nof, (ed.). John Wiley & Sons, New York, chap. 2.4.