Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Digital Manufacturing and RFID-Based Automation 49.4 Case Studies of Digital Manufacturing and RFID-Based Automation 875

Intra-supply chain information flow

Spinning

Optimization

enabling

technologies

Intra-supply

chain data

integration

tier

Information

systems tier

Data

acquisitions

tier

Other inter-supply chain

partners information network

Local IS

WMSOMS

Reader Tagged beam

Dyeing

Local IS

WMSERP

Reader Tagged beam

Wool buying

Local IS

WMSERP

Reader Tagged beam

EPCIS

Physical materials flow

Fig. 49.22 Architecture of RFID-

based ISCIS (IS – information

system, WMS – warehouse manage-

ment system)

3. Intra-supply-chain data integration platform: Fig-

ure 49.23 shows how the supply-chain information

is shared and integrated with the different supply-

chain parties. On this platform, an integrated

information system is built to enable the supply-

chain partners to download the information from the

chain. As discussed before, the RFID gateway can

Fiber production

Pallet level Pallet level

Reader

Reader Reader

Fiber dyeing

Carton level

Yarn spinning

Item level

Retail

Carton and item

level

Knitting and finishing

Fig. 49.23 Information sharing along the supply chain

act as a check-in and check-out system and synchro-

nize the internal information systems without delay.

Supply-chain partners can select the information to

be shared, which will be stored in a centralized

database.

4. Web-based platform: On this platform, a web-based

information-sharing portal is built so that supply-

Part F 49.4

876 Part F Industrial Automation

chain partners can retrieve related supply-chain

information. Intra-supply-chain partners can use the

web browser to search for information about the

chain. The production status, the supplier’s inven-

tory level, the delivery status, and retail sales data

can be displayed on the web-based platform. The

above platforms help an intra-supply-chain com-

pany to obtain data on their operations and provide

an interface permitting them to share information.

To facilitate the intra-supply-chain collaboration,

different optimization technologies, such as intelli-

gent agent and data mining, can be used to analyze

these data.

To realize the capability of the RFID-based ISCIS,

a case study was conducted in a reference site in the tex-

tile industry named Novetex. Novetex Spinners Limited

was established in 1976 and, employing over 2000 peo-

ple, is now recognized as the world’s largest single-site

woolen spinner. As part of the Novel Group, Novetex

has its headquarters in Hong Kong and a factory lo-

cated in Zhuhai, Southern China. The factory houses

four spinning mills, a dyeing mill, and 51 production

lines, resulting in an annual capacity of 7500t of high-

quality yarn. Novetex has recognized the need to be

a truly global supplier and has invested in ofﬁces and

agents worldwide to offer customers the most efﬁcient

service possible. The trend towards the establishment

of a global production network in the textile industry

poses some new challenges to Novetex; for example, ﬁ-

bres might be produced in one country, spun to yield

yarns in a second country, woven into fabrics in a third,

and sewn into clothing in yet another one, before being

ultimately sold somewhere else.

Novetex has traditionally used pen-and-paper-based

warehouse management processes. The recording of

the inventory, storage, order picking, packaging, and

stocktaking processes were heavily dependenton theef-

ﬁciency of warehouse operators, who used their own

judgment and perceptions to decide how, when, and

where to store the goods. This heavy reliance on pa-

perwork and on human input to update the inventory

data inevitably leads to data inaccuracy. This in turn

may affect the company’s decisions on inventory re-

plenishment, inventory control, and the ﬁrst-in ﬁrst-out

(FIFO) order-picking process. Heavy reliance on pen-

and-paper-based warehouse management processes has

limited the operational efﬁciency in inventory replen-

ishment, inventory control, and the FIFO order picking

process of Novetex. These issuesbecome morecomplex

in the era of the global production network.

As a result, the company has implementedan RFID-

enabled ISCIS. The outcomes of the implementation of

the system were positivein terms of increased stock vis-

ibility and data accuracy. The textile manufacturer can

synchronize the inventory status with the ERP software

without any delay. The attachment of an RFID tag to

each bundle of goods facilitates identiﬁcation and vi-

sualization of items. The storage location of the items

can be easily identiﬁed and retrieved from the ERP sys-

tem. Order picking and stock relocation is easier for the

warehouse.

49.4.6 WIP Tracking

RFID-based solutions have been used for a decade in

manufacturing. Their applications in a manufacturing

enterprise are usually tracking of parts during manu-

facture and tracking of assembled items, i.e., WIP.As

a result, efﬁciency can be increased while entry er-

rors and manpower are saved. For WIP tracking, the

tags can be attached to products as they are being as-

sembled or created along the production line [49.35].

The status of the product can be updated as it pro-

gresses along the production line via RFID readers

placed above and below its path [49.36]. According to

Hedgepeth [49.37], the history or route taken by an in-

dividual product can be ascertained from data stored

on tags attached to the item, by installing RFID read-

ers at single or multiple locations along the production

line. As a result, the location as well as the ﬂow his-

tory of any item can be recorded in the manufacturer’s

database. To realize the advantages of RFID-enabled

WIP tracking, a case study was conducted in a preci-

sion mold manufacturing company named Nypro Tool

Hong Kong Ltd.

Nypro Tool Hong Kong Ltd. is a precision injection

mold manufacturer which provides high-quality mold-

ing tools globally. Since the company provides products

with a wide variety, WIP items need to be transported to

different workshops for various processes such as ma-

chining, heat treatment, quality checking, etc. Each of

the items is accompanied by its unique drawing so that

operators can refer to its required operations. Currently,

Nypro is using a barcode system to monitor the ﬂow of

WIP items within the plants. For every processing oper-

ation of the WIP items, the status of the process must

be captured manually so that management can check

follow-up information for WIP through the database

system. However, the manual process induces a lot of

human errors and is time consuming. The application

of RFID in WIP tracking is believed to be able to au-

Part F 49.4

Digital Manufacturing and RFID-Based Automation 49.5 Conclusions 877

65 cm

70 cm

77 cm

50 cm

55 cm

File folder

RFID reader/

antenna

RFID tag

WIP location and process tracking

Configuration of the RFID system

Parking area

Fig. 49.24a,b RFID-based location tracking and process status tracking for WIP application (a) WIP location and

process tracking

(b) Conﬁguration of the RFID system

tomate the tracking process and reduce the number of

errors.

As shown in Fig.49.24a, which shows the loca-

tion tracking process, RFID gateways are set up at the

entrance of the workshops. WIP items and drawings

transported to the workshop arerequired to pass through

the gateway and the tag information for the arriving

item is automatically tracked when it is pulled through

the gateway. Figure 49.24b shows the setup for the

RFID-based process. The process is designed in such

a way as to capture the machining status of the WIP

items accurately and automatically. The result of the

trial implementation shows that the management of the

company can accurately plan job allocation during the

production process through automatic and accurate cap-

ture and tracking of the information related to the ﬂow,

location, and processing status of the WIP items. In

other words, visibilityof WIPinformation can be signif-

icantly enhanced and this makes manufacturing process

automation possible.

49.5 Conclusions

Digital manufacturing through virtual manufactur-

ing simulation, and real-time tracking of production

data in the supply chain, have led to improved ac-

curacy of information, reduction of human errors,

and automation of business operations. This chapter

presents a study of digital manufacturing which cov-

ers computer simulation of manufacturing processes,

and systems for manufacturing process automation

based on real-time online capture of RFID data from

the movement of materials, people, equipment, and

assets in the production value chain. Various case

studies are described which involve virtual assembly,

VP, PAM, warehouse management, WIP management,

and the management of a global production net-

work.

Virtual manufacturing has gone beyond the graph-

ical simulation of product and process design, has

accelerated the product development cycle, and deepens

the level of customer interaction in the preproduction

phase of products. Combined with the growing demand

for RFID technology for automatic capture, tracking,

and processing of the huge amount of data available in

the production of goods and services, digital manufac-

turing is going to revolutionize the way in which the

supply chain is managed. It will also greatly change the

behavior of global producers and consumers.

Part F 49.5

878 Part F Industrial Automation

References

49.1 H.C. Crabb: The Virtual Engineer: 21st Century Prod-

uct Development (American Society of Mechanical

Engineers, New York 1998)

49.2 W.B. Lee, H.C.W. Lau: Factory on demand: the

shaping of an agile production network, Int. J. Agil.

Manag. Syst. 1/2, 83–87 (1999)

49.3 RedEye: http://www.redeyerpm.com (2008)

49.4 Stratasys: http://www.stratsys.com (2008)

49.5 Autodesk: http://usa.autodestl.com (2008)

49.6 F. Kimura: Product and process modelling as a ker-

nel for virtual manufacturing environment, Ann.

CIRP 42, 147–150 (1993)

49.7 Lawrence Associates Inc. (Ed.): Virtual Manu-

facturing User Workshop, Tech. Rep. (Lawrence

Associates, Wellesley 1994)

49.8 C. Shukla, M. Vazquez, F.F. Chen: Virtual manufac-

turing: an overview, Comput. Ind. Eng. 13,79–82

(1996)

49.9 M. Onosato, K. Iwata: Development of a vir-

tual manufacturing system by integrating product

models and factory models, Ann. CIRP 42, 475–478

(1993)

49.10 K. Iwata, M. Onosato, K. Teramoto, S. Osaki:

A modelling and simulation architecture for vir-

tual manufacturing systems, Ann. CIRP 44, 399–402

(1995)

49.11 J. Gausemeier, O.V. Bohuszewicz, P. Ebbesmeyer,

M. Grafe: Cyberbikes-interactive visualization of

manufacturing processes in a virtual environment.

In: Globalization of Manufacturing in the Digital

Communications Era of the 21 Century-Innovation,

Agility, and the Virtual Enterprise, ed. by G. Jacucci,

G.J. Olling, K. Preiss, M.J. Wozny (Kluwer Academic,

Dordrecht 1998) pp. 413–424

49.12 K. Iwata, M. Onosato, K. Teramoto, S. Osaki: Virtual

manufacturing systems as advanced information

infrastructure for integrated manufacturing re-

sources and activities, Ann. CIRP 46, 335–338 (1997)

49.13 K.I. Lee, S.D. Noh: Virtual manufacturing system -

a test-bed of engineering activities, Ann. CIRP 46,

347–350 (1997)

49.14 S. Jayaram, H. Connacher, K. Lyons: Virtual assem-

bly using virtual reality techniques, Comput. Aided

Des. 28, 575–584 (1997)

49.15 R. Tesic, P. Banerjee: Exact collision detection us-

ing virtual objects in virtual reality modelling of

a manufacturing process, J. Manuf. Syst. 18,367–

376 (1999)

49.16 U. Jasnoch, R. Dohms, F.B. Schenke: Virtual en-

gineering in investment goods industry-potentials

and application concept. In: Globalization of Man-

ufacturing in the Digital Communications Era of

the 21 Century-Innovation, Agility, and the Virtual

Enterprise, ed. by G. Jacucci, G.J. Olling, K. Preiss,

M.J. Wozny (Kluwer Academic, Dordrecht 1998)

pp. 487–498

49.17 M. Weyrish, P. Drew: An interactive environment

for virtual manufacturing: the virtual workbench,

Comput. Ind. 38, 5–15 (1999)

49.18 RNCOS: RFID Industry – A Market Update,

http://www.rncos.com/Report/COM16.htm (2005)

49.19 R. Qiu, Q. Xu: Standardized shop ﬂoor automation:

An integration perspective, Proc. 14th Int. Conf.

Flexible Automation and Intelligent Manufacturing

(NRC Research Press 2004) pp. 1004–1012

49.20 C. Poirier, D. Mccollum: RFID Strategic Implementa-

tion and ROI: A Practical Roadmap to Success (Book

News, Portland 2006)

49.21 D. McFarlane, S. Sarma, J. Chirn, C. Wong,

K. Ashton: Auto-ID systems and intelligent man-

ufacturing control, Eng. Appl. Artif. Intell. 16,

365–376 (2003)

49.22 R. Qiu: A service-oriented integration framework

for semiconductor manufacturing systems, Int. J.

Manuf. Technol. Manage. 10, 177–191 (2007)

49.23 Kali Laboratories: RFID Implementation,

http://www.icegen.net/implement.htm (2004)

49.24 A. Kambil, J. Brooks: Auto-ID across the value

chain: from dramatic potential to greater efﬁciency

and proﬁt, White Paper of MIT Auto-ID Center (2002)

49.25 R. Qiu: RFID-enabled automation in support of fac-

tory integration, Robot. Comput. Integr. Manuf. 23,

677–683 (2007)

49.26 R. Moroz Ltd.: Understanding radio frequency

identiﬁcation, http://www.rmoroz.com/pdfs/

UNDERSTANDING%20RFID_November22_2004.pdf

(2004)

49.27 F. Klaus: RFID Handbook, 2nd edn. (Wiley, New York

2003)

49.28 B. Bacheldor: Aircraft Parts Maker Adds Tags

to Molds, http://www.rﬁdjournal.com/articleview/

2411/1/1 (2006)

49.29 C.F. Cheung, W.B. Lee: A framework of a virtual ma-

chining and inspection system for diamond turning

of precision optics, J. Mater. Proc. Technol. 119,

27–40 (2001)

49.30 W.B. Lee, J.G. Li, C.F. Cheung: Research on the de-

velopment of a virtual precision machining system,

Chin. J. Mech. Eng. 37, 68–73 (2001)

49.31 W.B. Lee, C.F. Cheung, J.G. Li: Applications of virtual

manufacturing in materials processing, J. Mater.

Proc. Technol. 113, 416–423 (2001)

49.32 Y.X. Yai, J.G. Li, W.B. Lee, C.F. Cheung, Z.J. Yuan:

VMMC: a test-bed for machining, Comput. Ind. 47,

255–268 (2002)

49.33 W.B. Lee, J.G. Li, C.F. Cheung: Development of

a virtual training workshop in ultra-precision ma-

chining, Int. J. Eng. Educ. 18, 584–596 (2002)

Part F 49

Digital Manufacturing and RFID-Based Automation References 879

49.34 I. Hipkin: Knowledge and IS implementation: case

studies in physical asset management, Int. J. Oper.

Prod. Manage. 21, 1358–1380 (2001)

49.35 Datamonitor: RFID in Manufacturing: The Race to

Radio-Tag is Heating Up in Manufacturing (Data-

monitor, New York 2005)

49.36 D.E. Brown: RFID Implementation (McCraw-Hill,

New York 2007)

49.37 W.O. Hedgepeth: RFID Metrics: Decision Making

Tools for Today’s Supply Chains (CRC, Boca Raton

2006)

Part F 49

“This page left intentionally blank.”

881

Flexible and P

50. Flexible and Precision Assembly

Brian Carlisle

Flexible assembly refers to an assembly system

that can build multiple similar products with lit-

tle or no reconﬁguration of the assembly system. It

can serve as a case study for some of the emerging

applications in ﬂexible automation. A truly ﬂexible

assembly system should include ﬂexible part feed-

ing, grasping, and ﬁxturing as well as a variety of

mating and fastening processes that can be quickly

added or deleted without costly engineering. There

is a limited science base for how to design ﬂexible

assembly systems in a manner that will yield pre-

dictable and reliable throughputs. The emergence

of geometric modeling systems (computer-aided

design, CAD) has enabled work in geometric rea-

soning in the last few years. Geometric models

have been applied in areas such as machine vi-

sion for object recognition, design and throughput

analysis of ﬂexible part feeders, and dynamic sim-

ulation of assembly stations and assembly lines.

Still lacking are useful techniques for automatic

model generation, planning, error representation,

and error recovery. Future software architectures

for ﬂexible automation should include geomet-

ric modeling and reasoning capabilities to support

autonomous, sensor-driven systems.

50.1 Flexible Assembly Automation ............... 881

50.1.1 Feeding Parts............................. 882

50.1.2 Grasping Parts ........................... 883

50.1.3 Flexible Fixturing ....................... 885

50.2 Small Parts........................................... 886

50.2.1 Aligning Small Parts.................... 886

50.2.2 Fastening Small Parts ................. 886

50.3 Automation Software Architecture ......... 887

50.3.1 Basic Control

and Procedural Features ............. 887

50.3.2 Coordinate System Manipulation.. 887

50.3.3 Sensor Interfaces

and Sensor Processing ................ 888

50.3.4 Communications Support

and Messaging........................... 888

50.3.5 Geometric Modeling ................... 888

50.3.6 Application Error Monitoring

and Branching........................... 889

50.3.7 Safety Features .......................... 889

50.3.8 Simulation and Planning ............ 889

50.3.9 Pooling Resources and Knowledge 890

50.4 Conclusions and Future Challenges......... 890

50.5 Further Reading ................................... 890

References .................................................. 890

50.1 Flexible Assembly Automation

Flexible assembly automation is used in many in-

dustries such as electronics, medical products, and

automotive, ranging in scale from disk drive to auto-

motive body assembly. The ubiquitous yardstick for

a ﬂexible assembly system is the human. While dedi-

cated machines can far exceed the speed of a human for

certain applications, for example, circuit board assem-

bly, there is no automation system than can assemble

a hard disk drive one day and an automobile fuel injec-

tor the next.

A ﬂexible assembly system must present, grasp,

mate, and attach parts to each other. It may also per-

form quality tests during these processes. In order to

be useful, it must compare favorably in terms of de-

velopment cost and product assembly cost with manual

alternatives. To date, automatic assembly has only been

justiﬁed for applications where volumes exceed tens of

thousands of assemblies per year.

Key capabilities in sensing, modeling, reasoning,

manipulation, and planning are still in their infancy for

Part F 50

882 Part F Industrial Automation

automated systems. The following discussion of ﬂexi-

ble automation for assembly will illustrate the current

state of the art. Further advances in ﬂexible automa-

tion will require machine control software architectures

that can integrate complex motion control, real-time

sensing, three-dimensional (3-D) modeling, 3-D model

generation from sensor data, automatic motion plan-

ning from task-level goals, and means to represent and

recover from errors.

50.1.1 Feeding Parts

Assembly requires picking up parts, orienting them,

and fastening them together. In almost all automatic as-

sembly systems in production today, parts are oriented

and located by some means before the assembly sys-

tem picks them up. In fact, it should be recognized that

orientation has value; if part orientation is lost, it costs

money to restore it. Due to this requirement for pre-

oriented parts, most automatic assembly systems only

deal with small parts, typically less than 10 cm

in vol-

ume. Traditional small part feeders include indexing

tape feeders (Fig.50.1), tray feeders, tube feeders, Gel

Pack (a sticky ﬁlm in a round frame for semiconductor

parts) feeders, and vibratory bowl feeders (Fig. 50.2).

Each of these feeders is typically limited to a fairly nar-

row class of parts; for example, tape feeders are sized

to the width of a part; a 3 mm-wide tape cannot feed

a 25mm-wide part. Tape and tray systems have sev-

eral disadvantages in addition to their cost: they are not

space efﬁcient for shipping as parts are stored at a low

density, and the packing material can add several cents

to the cost of each part. Larger parts are almost univer-

Fig. 50.1 Sticky-tape part feeder (courtesy of Pelican

Packaging Inc.)

sally transported on pallets, or in boxes or bins, and are

at best only partially oriented.

The development of machine vision systems has be-

gun to change this situation (Fig.50.3). Useful machine

vision systems began to emerge in the 1980s. Early

systems that could recognize the silhouette of a part

in two dimensions cost US$50000–100000. Today,

you can buy two-dimensional (2-D) vision systems for

a few thousand dollars, and the price continues to drop

rapidly. At the time of writing, 3-D vision systems are

beginning to emerge, and in a few cases, are being in-

stalled in factories to guide robots in the acquisition of

large heavy parts from pallets and even bins [50.1].

Vision systems are allowing the development of part

feeders that can separate parts from bulk, inspect the

parts for certain critical dimensions, and guide a robot

or other automatic assembly machine to acquire the part

reliably.

Several such part feeders are shown in Fig.50.4.

These feeders utilize either conveyors or vibratory

plates to separate parts from bulk and advance them un-

der a machine vision system, which then determines if

the part is in an orientation that can be picked up by the

assembly machine. If not, parts are usually recirculated,

since it takesa lot oftime topick them up and handthem

off to a part-reorienting mechanism, and then regrasp

them. These feeders can handle a wide variety of part

sizes and geometries without changing the feeder. In

fact, this class of feeder is increasingly being accepted

in applications where parts are changed daily or several

times a day.

Predicting the throughput of this type of feeder re-

quires predicting the probability that a part will come

Fig. 50.2 Bowl part feeder (courtesy of Pelican Packaging

Inc.)

Part F 50.1

Flexible and Precision Assembly 50.1 Flexible Assembly Automation 883

to rest in a desired stable state (will not topple or

roll) when it is separated from the other parts. The au-

thor worked with Ken Goldberg and others to develop

a method for predicting the distribution of stable states

from part geometry [50.2]. Part geometries can be com-

plex, and the distribution of stable states may not be

obvious. The part illustrated in Fig. 50.5, from a plastic

camera, has 12 stable states, of which four are shown.

Goldberg et al. developed an algorithm that could pre-

dict the distribution of these12 stable statesfrom a CAD

model of the part.

The ability to simulate feeder throughput from an

analysis of part geometry is an example of using mod-

eling and geometric reasoning to help design a ﬂexible

assembly system.

It is desirable to be able to generate both 2-D and

3-D vision recognition algorithms directly from CAD

models of parts. While this sounds simple for the case

of 2-D parts, to date the author is not aware of any

commercially available vision systems that offer this

capability. Real-world challenges from lighting, reﬂec-

tions, shadowing, lens and parallax distortion, camera

and lens geometry, etc. combine to make this a chal-

lenging task. The task becomes more challenging when

3-D vision is considered. Robust models that take into

account these factors need to be developed and exten-

sive algorithm testing done to ensure reliable real-world

vision system performance from CAD-generated algo-

rithms.

In summary, recent advances in machine vision are

allowing large parts to be picked directly from bins and

small parts to be picked from mechanisms that separate

them under a camera. These advances allow assembly

systems to be changed over quickly to new products

without designing and installing new part feeders.

50.1.2 Grasping Parts

In today’s commercial assembly systems almost 100%

of part grasping is done by either a vacuum cup or

a two-ﬁnger gripper with custom ﬁngers designed for

the particular part to be grasped. Even for systems with

the vision-based ﬂexible part feeders described above,

if we change parts, we must change the part gripper.

Part pick-up strategies are generated by explicit manual

training or programming. Obstacle avoidance of other

parts generally also requires part-speciﬁcprogramming.

This approach works ﬁne for systems that only have to

handle one or two types of parts for an extended period

of time.

Imaging electronics

Camera Camera

Material handling

Lighting

Software

Computer

Operator

console

Modem

Network

I/O interface

Lighting

LensLens

Imaging electronics

Fig. 50.3 Machine vision system (courtesy of High-Tech Digital,

Inc.)

As part dimensions shrink, new considerations enter

into gripper design. As part dimensions approach 1 mm,

chemical attraction,static electricityor otherforces may

exceed the force of gravity on the part. So, while it may

be possible to pick up a small part with a vacuum nee-

dle, it may be necessary to use air pressure or other

means to release it. With tiny parts, it can be very dif-

ﬁcult to grasp a part with the precision needed to place

it at the desired location. Therefore, for most very small

parts, machine vision isused to reﬁne the position of the

part in the gripper after it has been grasped.

a) b)

Fig. 50.4a,b Flexible part feeders. (a) Adept Technol-

ogy [50.3],

(b) Flexfactory [50.4]

Part F 50.1

884 Part F Industrial Automation

Fig. 50.5 Four of 12 stable states for a complex part

In order to maintain a stable grasp of a part, the

part should be uniquely constrained, with gripper me-

chanical tolerances taken into account; for example,

a two-ﬁnger gripper with an opposing V groove in each

ﬁnger would appear to locate a round part. However, if

the ﬁngers are not perfectly aligned, the part in fact will

roll between parallel plates of the opposite V grooves.

For this example, one V groove opposite a ﬂat ﬁnger

would be a better solution.

In a similar manner, attempting to grasp a prismatic

part with two ﬂat ﬁngers is a bad idea. Any slight vari-

ation in parallelism of the part or ﬁngers will result in

a poor grasp that may not resist accelerationor part mat-

ing forces. In this case three pads to deﬁne a plane on

one ﬁnger and a single pad to press the part against this

plane on the opposing ﬁnger is a better solution and

can handle draft angle on parts, ﬁnger splay, and other

tolerances.

These simple examples illustrate the more general

concept of planning and achieving stable grasp in mul-

tiple dimensions, and resisting slip and torque in the

dimensions important to the process.

In the popular image of a service robot, the robot

performs a wide range of menial tasks, perhaps cooking

and serving dinner, cleaning up after the kids or dust-

ing the shelves. Within the next few years it is likely

that machine vision technology will develop to the point

Fig. 50.6 A typical industrial gripper and several humanoid grippers [50.5]

where, given explicit models of objects, a robot could

pick up a large variety of objects from an unstructured

environment. However, this will be possible only if it

can grasp them.

Multiﬁnger prehensile hands were developed in re-

search laboratories in the 1980s. Several companies

now offer commercial versions [50.6, 7]. However, to

date, we lack the sensing and control technologies to

reorient a part within the hand once it is grasped, ex-

cept for ad hoc preprogrammed strategies developed

manually with a good deal of trial and error. Even the

automatic generation of stable grasping strategies for

these complex hands is still a topic for research and has

not found commercial acceptance.

What is often not realized by the casual observer

of these humanoid hands is that the control of these

hands is far more difﬁcult than the control of a six-axis

robot arm. Planning stable grasps with humanoid hands

from CAD models of objects has been the topic of nu-

merous research papers for many years now [50.5, 8]

(Fig.50.6). Kragic et al. [50.8] present a typical ap-

proach to this issue and point out that having a CAD

model and grasp plan are not sufﬁcient to achieve a sta-

ble grasp. The robot must also determine the actual

orientation of the object. Kragic et al. used simple ob-

jects with strong markers so that a simple 3-D vision

system could determine the object orientation for the

Part F 50.1