RFID and Agent-Based Computing

RFID and SAP: A Strategic Vision
Leading Edge Forum Technology Grant
Jonathan Gregory
Computer Sciences Corporation
May 2006

ii
Table of Contents
Preface...................................................................................................................1
Introduction...........................................................................................................1
Executive Summary .............................................................................................3
Business Case–Based Introduction to Basic Implementations of RFID ........4
Con-Agra ................................................................................................................4
Wal-Mart .................................................................................................................4
The Department of Defense ...................................................................................5
RFID Core Functionality.......................................................................................6
RFID Core Technology.........................................................................................8
RFID Tags ..............................................................................................................8
Unique Identification ...............................................................................................9
Bridging the Gap: The Divide Between the ERP World and the
RFID World ..........................................................................................................10
SAP’s RFID Solution: The Auto ID Infrastructure (Aii) Module......................12
Looking Forward: SAP NetWeaver and the Enterprise Services
Architecture.........................................................................................................13
Web 2.0 ................................................................................................................14
The Network Effect .............................................................................................16
Agile Systems .....................................................................................................16
Integrated Marine Multi-Agent C2 System (IMMACCS) ..................................17
Battle Scene from the Age of Empires .................................................................18
Asset-Oriented Computing................................................................................18
Revisiting the Marine’s Supply Chain ..............................................................19
Industrial Applications of the Asset Services Model......................................20
Recommendations .............................................................................................24
Appendix .............................................................................................................27
The Aircraft Engine Aftermarket ...........................................................................27
Details on SAP’s “Auto ID Infrastructure” (Aii) Product ........................................28
Complex Event Processing (CEP)........................................................................29
The Own Versus Influence Model .....................................................................31

1
Preface
Radio frequency identification (RFID) is an exciting technology that can be used
anywhere that a unique identification system is needed. It is used to track assets
from beer kegs to containers of spent nuclear fuel. It is used to track expensive
aircraft maintenance-oriented tools, and used to track library books. It is used by
Japanese school children to flag alerts if they encounter a problem when walking
to or from school, and used to allow cashless payments for services such as bus
transportation in skiing communities via a tagged ski pass. It can make existing
processes better, faster, and cheaper — or it may be used to provide totally new
sources of information and, in the process, enabling revolutionary new solutions to
problems.
The world of RFID has many new areas to write about and explore. This is due in
part to the foundational shift in computing that RFID represents. RFID offers
computing systems a more comprehensive means of gathering input about the real
world. Vast amounts of data may be continually collected without human
intervention, drastically lowering the cost of data collection. All kinds of assets may
now possess an identity enabling visibility into each particular asset’s life cycle.
New specialty areas and solutions emerge — RF security, data type standards,
sensor data transmission and standards, new protocols, new data processing
technologies, etc.
Tagged assets reveal the regular departure of physical assets from established
business rules or understandings of asset behavior. Information gathered from
RFID has shown consumer product managers that supply chains are much less
responsive than previously thought. Contractual obligations which previously were
assumed to have been kept are now realized to be noncompliant due to new audit
trails. The RFID technology reveals process inefficiencies, standards of
nonconformance, contractual breaches, and other flaws in asset management
processes. Thus, RFID becomes a catalyst for more rapid process improvements.
This document will attempt to synthesize the RFID subject into a rational means of
understanding the technology, the functionality, and the business implications.
After which point, this paper will also attempt to predict future enterprise resource
planning (ERP) and online technology and business trends, in order to describe an
ideal, futuristic vision for RFID adoption. Such a vision, though largely academic,
provides an end state that will help guide strategic goals, influencing present-day
adoption decisions. It is easy to “drink the Kool-Aid” when discussing the future of
RFID; I have attempted to stay grounded and build futuristic statements on solid,
credible, present-day references. Due to the varied interests of the readers of this
document, detailed technical comments have been largely compartmentalized in
the appendix.
Introduction
In the days of the automobile’s invention, people initially referred to the contraption
as a “horseless carriage.” They had to use the earlier technology to define the later
one. Such a characterization was accurate, but did not recognize the automobile’s
superior utility. In a similar form, the RFID technology has been referred to as
“wireless bar-coding.” This is an accurate depiction; however, it fails to recognize
the utility offered by the technology.
Small decisions at the beginning of an initiative can have enormous ripple effects
as those decisions are propagated. A rule of thumb is that each phase in the
product development cycle requires 10 times the work or cost to solve a problem
than in the prior phase. The purpose of this paper is to help identify design issues
with RFID solutions (specifically the technologies used to process the data streams
and derive value from the data) before they are created by offering information that
will lead the decision maker to visualize an optimal adoption of the various auto-ID

2
technologies given current expectations of the future of IT. Said differently, the
paper seeks to introduce core concepts surrounding RFID that will impact the
adoption of the technology for years to come. The paper will also offer a broad set
of principals for building a road map in order to achieve this vision.
This paper will introduce several business cases, followed by a basic functional
and technical review of RFID. The paper then turns to the subject of bridging the
gap between the RFID data collection world and ERP solutions. This will involve
information about present and future ERP solution characteristics, as well as the
characteristics of “new” technologies used to glean events from RFID data
streams. The future of ERP involves a service-oriented architecture, which
dovetails with “Web 2.0” concepts surrounding the next generation of Internet
solutions. The Web 2.0 model relies on and demonstrates the “Network Effect,”
and several examples of network-centric computing, including a model for asset
management, follow.

3
Executive Summary
RFID reduces the cost of data collection, enabling real-time asset data to drive
granular asset automation. The technology can be leveraged as part of an
infrastructure and computing model, which enables new asset management
solutions, or it may be limited to automation of existing processes, which can
provide very real (but limited) value.
Emerging and existing IT tools and methods will enable dramatic new business
models and applications. These tools include global standards, layered business
process management (BPM) technologies, object-oriented programming, instant
messaging, Web services, as well as new database and event processing models.
ERP investments will be utilized as a foundation for core business processes in
which specialists within the enterprise add local customization extensions for
localized process improvements.
The value of a network is determined by its size and the degree of contribution of
its component nodes. The “network effect” refers to Reed’s Law, which states that
the utility of a network increases exponentially with the number of nodes. This
network effect is seen in the phenomena labeled “Web 2.0,” which implements
tools that enable bodies of people to contribute to the benefit of the larger system.
An ideal RFID-enabled asset management solution will leverage the network effect
by empowering process owners distributed throughout an enterprise to contribute
to the overall value of the network. Emerging standardized computing models can
be made into simple toolsets that utilize standardized interface definitions to Web
and enterprise services.
RFID data collection employs massive data streams that require new IT-specific
toolsets; reveals new information about business activities; and offers information
for process improvement methodologies. This will pose new challenges for
organizational IT governance. The process by which business processes are
developed and modified must be redefined in light of the evolving distribution of
process authority. Furthermore, the existing ERP-driven business processes may
be extended by disparate business units as well as customized to an asset-specific
level.

4
Business Case–Based Introduction to Basic
Implementations of RFID
This section will speak to the very basic uses of RFID, from a business value
perspective. These basic implementations can provide tremendous value through
significant improvements in existing business processes.
Con-Agra
Due to an e-Coli threat in 2002, ConAgra was forced to recall 18.6 million pounds
of beef, which was more than 50 times the amount that they actually should have
had to recall (354,000 pounds). Con-Agra did not have a sufficient audit trail to
allow for the substantially smaller, more accurate, recall. The company
subsequently left that business.i
The use of RFID could have been used to track
cattle movement through the supply chain and thus limit the scope of the recall.
This instance of RFID is using it to record past events. The Con-Agra case
demonstrates the relevance of RFID for product recall purposes. In a similar
fashion, automobile tire makers are incorporating RFID tags into their product to
help facilitate the same such functionality in the event of a recall. These examples
demonstrate the simplest uses of RFID and how they can add value as a stop-gap
measure. Here, the data collected is used only for historical purposes, yet the
value proposition still justifies the investment. The next example implements RFID
as a means of measuring present conditions.
Wal-Mart
New York Investment House Sandford C. Bernstein & Co estimated that Wal-Mart
might save $8.35 billion in total annual savings when RFID is deployed throughout
its operations. The largest component of that savings is $6.7 billion saved in
automating manual processes by eliminating the need to have people scan bar
codes on pallets and cases, reducing labor costs by 15 percent.ii
“What Wal-Mart has done was simple. It used RFID data to add one key piece of
information: Do we have it in the backroom? It’s just a little tiny tweak — it didn’t
change the way it stocks shelves or the printout of the pick list — but the impact is
phenomenal.” iii
Bill Hardgrave
Director, University of Arkansas RFID Research Center
Wal-Mart is using RFID to depict present conditions, “Do we have it in the
backroom?” to drive automated reordering and receiving. According to the
University of Arkansas research, initial estimates of RFID reducing Wal-Mart out-
of-stocks at 16 percent were too low. The greatest improvement was seen in
goods that sold from six to 15 units per day. These items showed a 62 percent
decrease in out-of-stocks.iv
The Wal-Mart mandate and business case is very well known, as is the concept of
tagging items for recall. The dramatic value potential seems at odds with the
observations made by Gartner research vice-president for ERP, Brian Zrimsek, at
a symposium in Sydney, Australia, in November 2004. Mr. Zrimsek warned that
mere evolutionary enhancements, such as enhancing existing business processes,
are missing the power and potential that the RFID technology offers. His assertion
is that RFID is a revolutionary technology that is being sub-optimized to fit into
evolutionary projects. For example, the Wal-Mart dictate is a “tactical rather than
strategic” solution as it is only a better labeling systemv
. Mr. Zrimsek noted further
that the association between an event triggered by an RFID tag movement and a
corresponding business action, (termed “act-on-fact”) is a key RFID return on

5
investment (ROI) concept. A discussion about a truly revolutionary implementation
of RFID will be threaded throughout the remainder of this paper. The “Zrimsek
Hurdle” will be the moniker for gauging if a solution, indeed, is revolutionary. The
two elements of this hurdle are:
1. Create new business processes or operational models
2. The use of RFID data to enable automation (“act on fact”)
The Department of Defense
“Marines can see where critical items are, and that changes the dynamic… The
dialog (between forward operating bases and the logistic hub) has changed from
‘Where’s my stuff?’ to ‘Why isn’t my stuff moving?’ to ‘I want you to put my stuff on
the next truck because I can see it’s there.”vi
Alan Estevez
Assistant Deputy Undersecretary of Supply Chain Integration, Department of
Defense
Implementing RFID in the Marine supply chain has cut inventory value in the chain
from $127 million to $70 million. Average delivery times have dropped from 28
days to 16 days, the supply backlog of 92,000 shipments has fallen to 11,000.vii
In
this example, Marines can view the products in the supply chain of interest to
them, effectively viewing and influencing the future deliverable date of the product.
This differs from the Wal-Mart example as that solution uses a reorder point-
planning model that benefits from a stable supply chain, whereas the military
supply chain has greater logistical issues and variability.
It appears that part of the success in the Marine’s supply chain solution, based on
the above quote, is the inclusion of stakeholders in representing their own
interests. This is a very “Web 2.0” type concept (a network of interests providing
value symbiotically). Nobody is more keenly interested in getting the requested
supplies than the soldier making that request. That soldier is what this paper will
later refer to as the “local process owner.” A critical component to improving
processes is empowering local process owners with tools that they can use to
identify and solve problems on their own. More will be discussed on this topic later
in the paper.
These supply chain enhancements do not pass the Zrimsek Hurdle, as they do not
revolutionize business processes or provide “act-on-fact” automation; however, the
paradigm shift in empowering local process owners to gain visibility and “speak up”
for their assets is a significant step in the right direction (it “changed the dynamic”).
The act of noticing that a RFID-tagged pallet has not been moving, and in fact
needs to be shipped to a forward location, does not need to be performed by a
person. An IT system should be able to automate that functionality. The issue here
is that mature IT products that are truly asset- and local process authority-centric
are nonexistent. The “Web 2.0” computing model, as well as object-oriented
programming, offers a great deal of promise in this area.
As with any new technology, adopters require a strong ROI-based case to justify
the investment. Revolutionary or paradigm altering implementations of RFID are
difficult or impossible to build an ROI case for, given that the solution impact is not
fully understood. What was the ROI on the first business PCs? What is the ROI on
e-mail? Was the economic (and social) impact of the first intercontinental train or
first freeway or the Internet known before building it? Understandably, venture
capital and solution providers are focused on products that provide specific
solutions, most of which enable vertical integration in specific industries.

6
SAP is leveraging its ownership of business processes (as defined in its software
products), client base, and substantial size to provide dozens of industry-specific
vertical collaboration hubs for supply chain management (SCM) integration. RFID
is renewing interest in creating SCM hubs for vertical integration. What RFID will
not do is tackle the nontechnical reasons (organizational culture, power struggles,
etc.), which are the very human reasons for limited SCM integration and
automation. It is no surprise that overwhelmingly powerful customers (Wal-Mart
and the DoD) have blazed new trails in SCM integration via RFID. They are
organizations powerful enough to enforce a one-sided mandate on their suppliers.
As the benefits of RFID compliance are realized through the entire supply chain,
industry segments with more equally powered players will increase interest. An
example of this is the aerospace industry, with its supply chain of certified, high-
value items, with low lot sizes. This industry is one of several forming standards
groups hosted by the Electronic Product Code (EPC) global network. These
vertical solutions are likely not to pass the Zrimsek Hurdle, as they will automate
existing SCM-oriented business processes (not change existing processes);
however, they will enable “act of fact” automation.
“We do a lot of things now in transactional systems which are triggered by human
beings. Somebody inputting a customer order or a goods receipt and so on. This
can be automated. If an RFID tag just crosses a reader, it’s done.”
Claus Heinrich, SAP Executive Board Member
The ability of an ERP vendor to provide both elements of the Zrimsek Hurdle
(revolutionize business processes and “act-on-fact” automation”) will, by definition,
be impossible until the adoption of “service-oriented architecture” (SOA) or, in
SAP’s case, the adoption of the “enterprise services architecture” (ESA). Generally
speaking, the ESA is SAP’s implementation of SOA. This software architecture is
what enables an ERP system to allow affordable business process transformation.
These components will be defined and spoken to in more detail later in the paper.
This section discussed a few business cases and current adoption models of the
technology. The next section will speak to the basic functionality offered by RFID,
which empowers the above business cases.
RFID Core Functionality
The RFID technology provides a means of accurately and inexpensively collecting
real-world data. These new data streams offer information systems substantial
new sources of data input, allowing for dramatic new uses of information
technology to solve problems.
RFID offers three core asset management capabilities:
1. The means of uniquely identifying assets
2. The means of locating assets
3. The means of knowing the condition of assets (through the use of integrated
sensors)
The first core benefit of RFID is the ability to specifically identify assets. Enormous
amounts of asset-specific data are not captured, in part, due to the anonymous
nature of nonserialized or unidentified assets. Another reason is the difficulty of
collecting and integrating the data in an economical fashion. Both of these reasons
for not capturing much more asset-specific data are evaporating; RFID lowers the
cost of collecting data because it can happen automatically with no human
intervention. More data (about the past history, present environment, and future
anticipated use of an asset) may be collected “to” the asset, and the technologies

7
and business models of the “Web 2.0” phenomena demonstrate economical ways
of integrating data into services rendering value.
An RFID tag provides a unique number identifying each specific asset and a very
low cost mechanism for collecting data. Once an asset is identified, data may be
collected to that specific asset. Data can build into information with value as the
ecosystem of interests surrounding that asset consume and provide data specific
to that asset.
With RFID, one can ask and answer the following questions:
Using Identity
• Is there a relationship between a performance metric (e.g., product defects,
returns, supply chain velocity) and a possible cause or contributor to that
performance (e.g., raw material used, engineering change on subassemblies,
supplier, assembler, lot, tooling, worker, worker or tool certification)? This is
clearly a means of continuous improvement.
Incorporating Sensors
One patriot missile in a lot of 20 was mistakenly dropped. Handlers lost track of
which missile had been dropped, so all 20 missiles had to be shipped back to the
United States, overhauled, and returned at a cost of $21 million.
Further questions may be asked and answered when including sensors:
• Report what happened to an asset in transit (e.g., an aircraft engine or patriot
missile, shipped in a sea van container, is dropped from a damaging height; an
RF tag integrated with a shock sensor can record and communicate that
event)
• Is there a relationship between a performance metric and a sensor reading
(sensors include light, temperature, shock, vibration, motion, radiological, call
buttons to signify events)?
• Report what is happening right now (e.g., a motion-activated tag transmits
while the plant is closed, raising a theft alert)
• Warn what might happen (e.g., a tag with a temperature sensor detects a
temperature that will destroy perishable goods, which raises an appropriate
alert)
Leveraging Location
By knowing the location of an object (and the time of the location read) one can
answer the following questions:
• Where is the asset right now (an obvious, but highly useful function)
• Is the asset where it is supposed to be, according to its workflow system?
• Is the asset combined with other assets that break a business rule or could
pose a hazard (e.g., environmental health and safety (EH&S) risks or
regulations that require chemicals to be separated)
In preparing for the War in Iraq, a sea van container carrying medical supplies was
erroneously removed from its vessel at a port in Hong Kong. Before the vessel
could leave, the application monitoring the RFID tag on the sea van container
alerted military logistical personnel, who were able to get the container placed
back on the vessel before it left for the Middle East. It would have taken months to
have corrected the distribution error.

8
Finally, this data can be incorporated into automation solutions that perform the
following:
• As soon as a particular asset is detected as having arrived at the dock door, a
certified worker is notified to process it for a rush order
• Predict a bottleneck in operations before it occurs and take corrective action
• Combinations of chemicals that are not supposed to be transported together
are mistakenly placed on the same truck, which approaches a gate to exit a
plant location. The RFID data triggers an alert and a security gate is prevented
from opening
• When any condition or situation a (knowledge) worker can think of occurs, do
whatever the (knowledge) worker intends or instructs. These conditions and
actions may be served by Web or enterprise services
The RFID technology, combined with robust systems and applications, can be a
component to moving the “tense” of business management of assets from past
tense (what happened), to present tense (what’s happening now), to the future
tense (what’s about to happen).
RFID Core Technology
This section will offer a basic introduction to RFID technology. An RFID system is
composed of tags, which are placed on assets and transmit unique data to readers
(also called access points) via an “RF” (radio frequency) signal. The access points
transmit data to core IT systems, which process the data and possibly interface
with external networks that contain further data about the tagged asset. This is
similar to how a garage door opener works and is seen in current everyday
activities such as the E-Z Pass automated toll collection, key-card entries to
buildings or rooms, and store security tags used on items such as clothing to
prevent shoplifting.
RFID Tags
RFID technology can be broken down into the categories of “active” tags (aRFID),
which are battery powered and have a longer read range (up to several
kilometers), and “passive” tags (pRFID), which cost less, have no battery, and
have a shorter read range (usually under 12 feet). An active tag may be used to
determine an asset’s location, even when the asset is not in close proximity to a
reader, whereas a passive tag’s location is typically only known when the tag
passes by a fixed gate. For this reason, it is important to distinguish between
active and passive tags from a continuous data perspective. An active tag in range
of an RF access point will continually generate location-based information.
Inferences may be drawn between the location of such an asset and the machine
used to finish it, the tooling used to calibrate it, the employee performing the work,
the supplier of the subcomponent or raw material, etc. Tag frequencies, sizes, form
factors and costs vary widely. There are also hybrid tags that include sensors of
various kinds, which may or may not be battery powered.
Passive RFID tags, given their short-range capabilities, are typically collected upon
movement through a gate — such as a loading dock doorway. Asset identity and
location is collected automatically and without human error. It is important to note
that the passive technology encounters real physical limitations when in
environments unfriendly to electromagnetic fields. Dense liquids and metals can
disrupt RF communication and limit the readability of tags located in the middle of
a pallet. Technology improvements (such as ultra-wide band communication) and
process solutions (such as storing all pallet content on a single pallet-connected
active tag) work around these limits of physics. Note that the RFID communication

9
solution is dependent on the data collection environment, which will result in varied
infrastructure costs.
Active RFID tags offer a larger read range, and therefore do not require a gate to
pass through for collecting data. While a gate on a passive tag collects asset
location at a particular time, active solutions allow for continuous data streams
denoting asset location. As with passive tags, active tag communication to access
points may be impacted by environmental factors. A tag in an outdoor environment
will encounter less interference than a tag in an indoor environment; say on a shop
floor with large heavy machinery running. Location-based mechanisms include
TDOA (time delay on arrival) and signal strength–based mechanisms. The TDOA
method measures the amount of time that it takes for an active tag’s signal to be
captured by three access points. With that information, the access points can
triangulate the asset location. This method encounters an issue when interference
delays a signal so as to misrepresent the distance of the tag from one or more of
the readers. The interference of a particular indoor location will increase the
density of access points required to make accurate, location-based reads. Outdoor
environments enjoy much less interference and add the capability of integrating
GPS-based location mechanisms into the RF tags.
A portable deployment kit provided by Savi Technology provides a mobile RFID
solution that extends the military supply chain into expeditionary operations where
there is no established RF infrastructure. This solution connects portable RF
readers (access points) to core systems via iridium satellite links. Initial units are
being deployed to Southwest Asia.viii
DoD is also testing RFID tags (dubbed the third-generation RFID with satellite
communications or “3G RFID w/SATCOM”) that incorporate GPS location sensing
and satellite communication to enable global tag location (within 3.5 feet) at any
time.ix x
The 3G RFID w/ SATCOM tags offer an RFID “tag” that is globally visible via
satellite. There is no reader or local infrastructure required to interrogate the tag.
While the “tag” cost will surely be high, in time the costs of all technologies
decrease due to the cycles of consumption volumes that drive mass production,
which lowers costs and spawns more consumption. Is this the future of closed-loop
use of RFID?
Unique Identification
A fundamental concept of RFID is the notion of unique identification. If a bar-coded
item is scanned twice, the system will reflect the existence of two items; the
system failed to recognize that the same asset was scanned twice. RFID tags
have a higher data capacity which allows for each item to be serialized; so if the
item is scanned multiple times in an RFID-enabled system, the system will
recognize the dual scans and filter out the erroneous data. Another advantage of
an RF tag’s greater data capacity is the ability of RF tags to transmit self-
describing data. For example, a tag can describe what the data is that it is
transmitting. It can indicate that it is transmitting a VIN number with the number
value, whereas a typical barcode only holds a value. Typically, the context of the
bar code directs the person scanning the item to properly classify the data.
For example, a worker at a receiving dock uses a bar code scanner to populate
fields in an application running on a fixed or handheld PC. The worker determines
which application to run, which field in the application to capture the data to, and
the worker makes judgments on what to scan. If this worker saw a package of
Doritos brand chips from someone’s lunch pail, they would know better than to try
and scan that bar code into their application. The system relies on the worker’s
judgment, which invites some degree of risk as people are fallible. People can

10
scan the same items twice, or miss items. If the Doritos chips had an RF tag, the
item would be possibly be read by an RF system; however, the self-describing
data (indicating the grocers’ data standards) would tell the system that the
scanned item was not relevant to the application.
The EPC global data standard for RF tags includes a product classification code
(which is defined by a global repository of product codes) and serial number for
uniquely identifying the tagged object. The product classification code may be
utilized to filter data streams (for example, filtering out the Doritos tag data) for
relevant tags. It may also be leveraged in process control. An IT system can use
the EPC standardized data, (and the global product code repository) to know what
kinds of products it is reading, even if it does not have data about the particular
serial numbers of the assets being read. This could be used by a system to
validate asset combinations. For example, medical drug combinations could be
checked automatically for violations of established medical rules; subassemblies
on system-level products such as jet engines or automobiles validated in the
manufacturing or aftermarket space; sea van containers of goods automatically
scanned to apply export restrictions, tariffs, or other rules.
Bridging the Gap: The Divide Between the ERP World and
the RFID World
The Wal-Mart example previously noted offered a potential savings of over $8
billion; however, the solution is estimated to collect over 7 terabytes of pRFID data
per day once fully implemented. In comparison, a large ERP transactional system
holds about 5 terabytes of data. The Library of Congress contains about 20
terabytes. The enormous volumes of continuously flowing RFID data requires new
solutions that employ event filtering and distribute business logic that previously
would have been stored on ERP or other core computing systems. The “traditional”
model of storing data to disk for later batch processing does not work in this
environment.
Database “on disk”
Structured query language
Database “in memory”
Event processing language
Differences in Processing
Differences in Data:
Entity level data
Manageable volumes
Long-term data retention
Low level, raw data
Large volumes
Temporal data
RFID and Auto ID
Environment
The ERP data and processing model, depicted on the left side of the above
diagram, is disk-based. It utilizes relational databases to manipulate manageable
volumes of data that are retained for the long term. Data entry is highly refined
(e.g., input a sales order line item for a particular company code) and requires a
trained person to input much of the data. Data is processed from the disk, or
database cache, utilizing a form of structured query language (SQL). SQL
leverages relational database concepts to join tables of information with “key”
linkages.

11
The right side of the figure depicts the RFID environment. This environment has
automated streams of data continually flowing from RFID readers. The data is
unrefined, often transmitted in bursts, and (as the Wal-Mart example showed)
provides much more data than typical IT systems are designed to process. This
environment can demand the use of “In Memory” databases1
.
There are several reasons to use memory instead of disk when processing
streams of data. Memory is about a thousand times faster than disk, enabling
better throughput. Also, the data is not necessarily useful after a few moments, so
storing to disk adds no value for the long term. Solutions implemented for the ultra-
high volumes of streaming data found in financial securities trading show an even
more highly performing means of processing streams of data. These solutions
bypass memory and send data streams directly to the processor (roughly similar to
how a network router works).
In place of SQL, event processing language (EPL) may be used to answer
questions like “tell me when these two events occur within 30 seconds” — or,
better yet, “tell me when these two events do not occur within 30 seconds.” SQL is
not built to answer questions such as this. The EPL construct also relates to a
technology concept known as complex event processing (CEP), which tracks
timing and causality of activities within the IT landscape. CEP is a very new
concept that can also provide system health and security functionality (see the
appendix entry on CEP for more details).
A good primitive example of an in-memory database is a Microsoft Excel
spreadsheet. As each cell is updated, any related calculations occur immediately.
The cell data is processed in real time and is not stored on disk (until the user
decides to save the spreadsheet). There is no disk-based computing in Microsoft
Excel.
In-Memory Database Excel Example
Asset ID Pedigree Location (X,Y) Zone Alert Console
0000001 Dynamite 1, 1 1 ALERT! Dynamite Near
Welder!
0000002 Toolkit 4, 3 2
0000003 Extinguisher 7, 2 3
0000004 Welder 3, 1 1
0000005 Truck 5, 3 2
0000006 Safety Kit 5, 3 2
Rule Definition Zone Definition
Dynamite Can’t work
with
Welder Zone 1 is where x is between
1–3
Welder Requires Extinguisher Zone 2 is where x is between
3–6
Truck Requires Safety Kit Zone 3 is where x is between
6–9
Here is a spreadsheet depicting the application of an EH&S rule that is designed to
prevent situations where two active tagged items (Dynamite and Welder) should
not be in the same zone. The X, Y values depict grid locations provided by a
location appliance that processes TDOA, signal strength, or GPS-based location
data. The X, Y values are continually updated in real time, the Excel example
translates the X, Y data into zones, and pedigree-based rules are built to raise
1
Also known as a “Real-time In-memory Event-drive Database” (RIED). Gartner projects
that commercial RIED solutions will grow significantly in the next 2 – 5 years

12
alerts when zone violations occur. Here the EPC global-provided pedigree
standard can be used.
The above in-memory database example demonstrates real-time processing of
multiple variables (e.g., location of dynamite boxes and location of welders). These
databases could process large numbers of complex rules in real time, which this
paper will refer to as “multi-variable” data processing. An important distinction
needs to be made between multi-variable data processing and “serial” (or single
variable) data processing. Traditional solution platforms enable scalable serial data
processing solutions (SAP’s Auto ID Infrastructure (Aii) is an example, which this
paper will discuss). These solutions process one record of data at a time, limiting
the ability of the system to scale applications such as the above Excel-based alert
utility.
SAP’s RFID Solution: The Auto ID Infrastructure (Aii)
Module
Given their control over centralized business processes, SAP has an enormous
competitive advantage when selling RFID middleware to its installed base. SAP
offers Aii as a middleware package that provides several out-of-the-box automated
functions as well as a platform (built on the same basis structure as R/3) for
solution development that includes the ABAP and Java development
environments.
Two flavors of the latest Aii version (4.0) offer either a standalone solution, which
requires 2 “custom” interfaces, or an integrated solution, which offers connections
into “XI” (eXchange Infrastructure: SAP’s EAI module), EM (Event Manager), and
R/3 (transactional system). The integrated solution includes R/3 core code
modifications that integrate with the out-of-the-box Aii solution. The Aii solution
may be a central instance located in the data center housing the R/3 instance, or it
may be a distributed solution, allowing for multiple Aii instances in the various
divisions of a corporation.
Aii offers value in the extension of standard SAP business processes (as defined
in the R/3 transactional system) to incorporate RFID data collection. It is highly
likely that SAP will continue to develop RFID process extensions in Aii, as well as
incorporate Aii into the dozens of vertical industry SCM hubs it is developing.
There is also value in incorporating standard Aii functionality into existing mobile
applications served by SAP. The out-of-the-box Aii process extensions apply only
to standard SAP business processes. For example, customized shipping solutions
in SAP will prevent the enterprise from utilizing the out-of-the-box shipping solution
built in Aii. The value proposition of the Aii functionality, therefore, is highly
dependant on the degree of process customization made in the R/3 product.
A potential benefit of Aii is the development platform that it provides. This is the
same basis platform, NetWeaver, upon which other SAP components are built.
The underlying basis technology at the core of the Application Server is one of
serial (not multivariable) data analysis. (The Aii module processes one record at a
time, and is not designed to analyze multiple variables in memory. SAP could
possibly leverage the Live Cache technology used in its Advanced Planning and
Optimization module to provide such in-memory functionality.) This poses an
acceptable limitation on the functionality offered by Aii, as that functionality is
limited to serial process extensions of core R/3 business processes. If needed, an
in-memory database could be stood up easily between Aii and RFID readers. That
database would be able to capture SAP-relevant events and pass them on to Aii,
XI, or even R/3 for processing. Further details on this and other elements of Aii are
available in the appendix.

13
Looking Forward: SAP NetWeaver and the Enterprise
Services Architecture
In the late 1990s, ERP vendors found customer dissatisfaction with their products
in the area of enabling business process change. The companywide
standardization of data and processes (which consolidated numerous legacy
applications) that made ERP valuable to large corporations brought with it a
monolithic system that raised barriers to business process change. Process or
data changes intended to suit one division may impact other divisions. This was
the seedling for the present ESA initiative. The ESA is SAP’s implementation of the
SOA. The ESA breaks down enterprise IT functionality into component services
that may be inexpensively organized into applications with BPM tools. This allows
different divisions to customize applications without impacting other portions of the
organization. The BPM tools interface is much like Visio, except the blocks in the
resulting diagram are programmatically linked to services with standardized
interfaces. An example to help in understanding this follows.
A general contractor decides to build a division of houses. He opens the yellow
pages, looks up “services” such as roofers, foundation contractors, framers,
financers, realtors, landscapers, etc. The general contractor organizes these
component services in such a way as to build and sell houses. In the same way,
autonomous enterprise services are found in a kind of yellow pages directory that
have standard interface definitions, organized using BPM tools to build
applications.
An example of the usage of this computing model is found at a large military
aircraft manufacturer, which implemented an Intalio BPM solution to streamline
spare parts order validations. The problem they had was that aircraft spare parts
orders would be validated in SAP one error at a time (SAP would kick out the sales
order on the first error, so if there were 20 errors in the order, the system would
kick the sales order out 20 times for correction before being able to finally process
the order). A solution that would have required 3,000 person days to build in SAP,
took 300 person days to build with a BPM solution that employed SAP function
modules and programs as services. This solution cost less to maintain than an
internal ERP customization, which would require additional work (validating and
adapting the solution) for every ERP software upgrade. The BPM solution resides
outside of SAP and, therefore, avoids the costs associated with upgrading custom
ERP applications.
Several BPM tools, technologies, and standards have been developed in
anticipation of a steady increase in SOA/ESA-enabled solutions. SAP and IBM
coauthored a position paper called “BPEL for People,” which builds upon the
established “Business Processes Execution Language (BPEL)” standard to add
components that programmatically define how business processes interact with
applications users. Given the marketing attention and investment statements made
by SAP, it appears that emerging standards (such as BPEL) will be incorporated
into the NetWeaver/ESA solution set to offer an internal, intranet-based BPM tools
(in fact, the XI EAI tool offered by SAP is BPEL-compliant). SAP is leading the way
in Gartner’s “magic quadrant,” which measures vision and ability to execute.
The SAP Aii solution does not initially pass the Zrimsek Hurdle as it, by definition,
does not create new business processes; rather Aii’s role is to extend and
automate existing R/3 business processes. However, the Aii development platform
offered may be used to define and automate new solutions. The ESA introduces
the ability to rapidly change and customize business processes, which opens the
door to change ERP from a monolithic implementation of standard processes to a
solution that allows specialty areas (e.g., the above example’s military spare parts
order fulfillment department) within the organization to make the system work for

14
them. When predicting the optimal use of RFID in ERP, it is important to anticipate
the future ERP environment. The ESA model offers a means of building adaptive
business processes and a means of integrating functionality from external
infrastructures (including Internet services and EPC global network data) into
business processes and product offerings. It is this environment that RFID will
mature in, as well as the larger Web 2.0 environment.
Web 2.0
The Web 2.0 concept is a broad recognition of the changing nature of the online
experience. It is partially enabled by the same concepts as an ESA or SOA. The
Internet is becoming increasingly personalized and participatory for an increasingly
connected population. Portable devices abound with new experiences for the
mobile user, and new avenues for marketers to establish direct relationships with
consumers.
The cell phone has become a new transaction platform. A cell phone camera can
be used to take a picture of a “spot code,” a symbol which is interpreted and
resolved into a URL or Web service. The spot code shown below takes the user to
CSC’s Leading Edge Forum Web site. Likewise, a universal identifier (UID — a
two-dimensional part marking that is globally unique2
) could be interpreted and
resolved into a Web service that can provide services specific to the item scanned
(for example, take a picture of a spot-code or UID on a copier machine to access a
Web service that collects or displays maintenance history data, service manuals,
or provides a service to reorder supplies). With a cell phone, a marketer can gain
access to customer identity (thorough the cell phone account), preferences and
purchasing history (through the customer’s online profile), where they are (via
location tracking services on cell phones), and what service or even what specific
item they are interested in (via scanning of spot code, UID, or RFID scan, which is
a feature in some cell phones used in Asia). Products need not be located in
stores for whimsical consumers to scan and purchase on demand at any location.
Spot Code Universal Identifier (UID)
Google has indicated investment in a service that would allow a consumer to
compare prices in real time. Using this service, a consumer takes a picture of the
bar code on an item (presumably in a store, but it could be a purchased product as
well), the service looks up and provides a list of similar items, with their sale prices
in proximity to the user, and offers directions to each location. The next step could
be to allow online price comparisons and ordering of the same product. Google
CEO Eric Schmidt has stated that he wishes to connect the world’s buyers and
sellers. Surely RFID plays a large role in this goal.
2
The DoD estimates that there will be 100 million parts marked with UIDs by 2010.

15
Speaking on how Google’s business model relates to RFID:
“I think it’s interesting to think about what happens when you know where
everything is…with RFID we have the ability to know where a good portion of the
physical things in the world are and what happens to business models then…it’s
going to change the way we do business”xi
Dr. John R. Williams
Director MIT AutoID Lab
(Selected as one of the 50 most powerful people in networking)
The comment by Dr. Williams suggests that asset location, which will be afforded
by RFID, will become a publicly accessible set of data that may be utilized by
solution providers such as Google. Such an intersection between asset location
data and the rich functionality of Internet solutions offers a paradigm change open
to new business models.
The next wave of Internet-based solutions (Web 2.0) demonstrates how cheaply
and easily data can be captured and computed at very large scales. The Web 2.0
model also shows that the resulting computed information is consumed in very
small units. For example, Google AdWords services a $6 billion industry served at
about 50 cents a shot. The zillow.com site will provide comparative market
analysis real estate estimates anywhere in the nation instantaneously and for free.
“The broad and rich foundation of the Internet will unleash a ‘services wave’ of
applications and experiences available instantly over the Internet to millions of
users…This coming ‘services wave’ will be very disruptive…The next sea change
is upon us. “xii xiii
William H. Gates
Chairman and Chief Software Architect, Microsoft Corporation
Apparently Microsoft is taking this sea change seriously, as the corporation
recently adjusted projected development spending in fiscal year 2007 to be $2
billion more than expected. “It sounds like you’re building a Google or a Yahoo
inside the company,” said Goldman Sachs & Co. analyst Rick Sherlund to
Microsoft CFO Chris Liddell.xiv
Web 2.0 demonstrates a few noteworthy points:
• Technology is becoming a forum for expressing a person’s identity
• Solutions offer focused information drawn from large amounts of data
• Service performance and ease of use is a critical success factor
• Software as a service, either for free (with targeted advertising providing the
funding for the business model) or for a small charge
• Increased ease of online payment for services
Not all of the processed data is consumed. Zillow.com presumably has every
address in the United States computed; surely not every address will be
requested. In addition:
• Solutions may be utilized as component services in an SOA, which leverages
standard and inexpensive integration toolsets
• Users are transacting and contributing online in new ways, with new devices,
(note the cell phone) with greater degrees of personalization

16
The Network Effect
David P. Reed asserted that the utility of large networks can scale exponentially
with the size of the network. This “network effect” is canonized as Reed’s Law.
This utility is demonstrated in the number of users of telephones. If only a few
people have a telephone, the utility of having one is much less. The example of the
Marine’s supply chain improvements relies on a network of interested parties
speaking up for their shipments. Ebay relies on networks of buyers and sellers
working through its online auction site. Value is derived from networks, and is
increasingly valuable when the network allows and enables contribution by nodes
(network participants). The network effect is a large component of the success of
Web 2.0. Some Web 2.0 models rely on high consumer volumes, providing small
bits of value, to justify large IT investment solutions (e.g., Google AdWords). Other
Web 2.0 solutions rely on groups that do more than consume information, but
contribute to the overall value of the network (e.g., “ebay” or “experts exchange”).
An optimal infrastructure for asset management will offer a means of knowledge
worker contribution in such a manner.
This concept is built upon by the popular book The Wisdom of Crowds by James
Surowiecki. Crowds can make very unwise decisions when each individual is
attempting to conform to a perceived consensus; this is titled “Groupthink”xv
. Mr.
Surowiecki states that the key qualities of making crowds (any group of people
acting collectively) smart are diverse, decentralized, and independent nodes
contributing to a mechanism capable of summarizing the group’s opinions into one
collective verdict.xvi
An example of this is a betting pool for a horse race or sporting
event.
The network effect has been largely associated with social networks. The SOA
seeks to lower the barriers between computing services to enable easier adoption
into networks, leveraging the network affect. An example of the network effect
where the nodes are computing agents, instead of humans, is found in agile
systems.
Agile Systems
An agile system is defined as one that is able to rapidly adapt to changes in its
environment. Such a system is composed of a large number of small, loosely
coupled “agents” or software objects performing specific calculations on data
streams and passing on computed results to accomplish an overall objective.
Examples are found in military systems that guide a missile to a particular target
and in financial systems turning profits from arbitrage trades between several stock
markets.xvii
The figure that follows illustrates a variation on agile systems that
associates software agents to real-world items and dedicates agents to warfighters
as well.

17
Integrated Marine Multi-Agent C2 System (IMMACCS)xviii xix
The IMMACCS assigns a software agent to a warfighter once that individual’s
location is known (for example, when they power on their GPS-enabled palm pilot).
That software agent is dedicated to the interests of that particular warfighter. It
constantly looks for all available intelligence on known enemy locations within a
certain perimeter (represented by the circle surrounding Alpha Company) and
notifies the warfighter. In urban warfare environments intelligence is supplied, in
part, by the network of warfighters inputting known enemy combatant locations in
real time.
Somewhat similar examples of these systems are found in computer games which
leverage object-oriented computing to produce highly complex simulations of a
“real- world” environment. The real-time, strategy-oriented computer game, “Age of
Empires” produced by Ensemble Studios and published by Microsoft in 1997,
depicts a life-like environment with populations of civilian and military characters
that the gamer manages in order to build both an economy and a military
component. Networked players can share the same landscape, or game board,
while they manage their particular tribe’s economic and military growth in order to
wage war to dominate a particular landscape.
The Age of Empires (AoE) game is a great example of object-oriented (OO)
programming. OO programming encapsulates data and methods into a single
program, which may be “instantiated” (manifested, incarnated, turned into an
actively running program) multiple times. An OO program is like a cookie cutter.
Once it is defined it can stamp out multiple cookies (actively running programs)
that are independent of one another (The cookie dough is memory with a dash of
processor cycles). Once created, the cookies have no relationship to one another;
they can be modified (decorated) but will retain the “inherited” shape given by the
cookie cutter.

18
Battle Scene from the Age of Empiresxx
Looking at the AoE game, each individual unit on the “game board” is an object. A
character has set capabilities (e.g., can build buildings, fish, chop wood, move
around, and has workflow rules it follows) and variables (e.g., health points, attack-
and-defend variables, speed of movement, and inventory of collected raw
materials). Dozens of characters of different types are guided by the gamer, each
character an instantiated object representing an independent program. Changes to
one character have no impact on other characters, yet all objects can interact with
one another in an “arms-length” fashion.
This general concept is in contrast to the ERP environment, where any change to
a data structure or process has a broad impact. If a developer wanted to change a
material master record in an ERP system so that just one record had a field that
was longer, they could not do so without impacting all the records in that table.
Likewise, if one wished to change a process, such as the sales order process, for
just one sales order, this could not be done without impacting all sales orders
flowing through that process.
Asset-Oriented Computing
The future of ERP is ESA, component services organized by BPM tools for diverse
applications. The future of the Internet is Web 2.0, cheap and valuable information,
personalized to user preferences, and even a mobile users’ location, provided on
demand and programmatically consumable. The cost of customizing business
processes drops and IT functionality and information available to workers
increases, both from internal enterprise services and from external Internet-
provided services.
The network effect relies on nodes of self-interest providing value in a symbiotic
fashion, best leveraged when individuals are empowered in their capabilities to

19
develop solutions. How might the network effect be realized with a corporation’s
physical assets? A concept for consideration is to associate one instantiated object
with a single RFID tagged asset (product or non-product) or a logical asset
collection (such as items on a pallet, a kit, or a lot).
It is at this point that many of the concepts previously described come together to
form a unique concept that this paper will refer to as “asset services.” This concept
is highly academic in nature3
, but provides an end-point vision for asset
management solutions. This concept departs substantially from conventional data-
processing solutions and proposes a platform for enabling a “network effect” for
asset management. The key to the network effect is the participation of the nodes
on the network. The model seeks to provide knowledge workers with toolsets that
enable the customization of asset “behavior,” thus exposing asset management to
the network effect.
Recapping several concepts:
• RFID provides the fundamental capability to uniquely identify assets in a
manner providing a great deal of data at a very low cost of data collection.
• The network effect is seen in many Web 2.0 innovative services. The value of
the network is in the size of the participating nodes, per Reed’s Law.
• The network effect was exemplified in the Marine’s supply chain
improvements, which enabled interested parties to view and affect the location
of in-transit assets.
• Both the Web 2.0 and ERP worlds are engaged in building reusable services
with standardized interfaces. This lowers the barriers to entry for solution
components to play a part in a larger network effect.
• BPM tools are used to organize services into tailored applications and offer a
means of programmatically defining business processes.
• Object-oriented technology has specifically demonstrated capabilities relevant
to the asset services model both in highly advanced applications (such as
IMMACCS), as well as more commonplace applications such as the Age of
Empires game. The object allows inheritance of standard behavior, but also
allows customization per object with no negative impact on the overall system.
These concepts can be formulated together into a strategic infrastructure for the
management of assets in a setting where change is constant and the process
authority is either localized or specialized. The IMMACCS dedicates a software
agent to the warfighter which looks out for the warfighter’s interests; why not have
a dedicated software agent for an asset that does the same?
Revisiting the Marine’s Supply Chain
As previously noted, the success of the Marine’s supply chain improvements
stemmed from forward commanders being able to view real time, in-transit asset
locations and act upon that information. The forward commander’s checks and
phone calls for expedition could be replaced with dedicated agents, which the
forward commanders manage. This would be conceptually similar (but much more
robust) to an e-mail message that one would flag for a reminder (the e-mail
message is singled out and customized to serve the readers’ agenda, likewise an
3
Note that the author has visited a startup company with several of the components in place
for this computing model, so the model may become a strong commercial tool in years to
come. Any implementation of this model would rely on several integrated (and standards-
based) IT infrastructure- oriented products, not on one standalone product.

20
asset service may be singled out and customized with a similar degree of ease).
Such agents could do much more than monitor deliveries, as the next few
examples demonstrate.
The DoD is working to incorporate RFID with sensors in munitions in order to track
life- cycle data on missiles. For example, a lot of 200 sparrow missiles procured in
1980 was broken into two groups; one group was stored on the USS Enterprise
(posing a potential salination-based corrosion), the other in Saudi Arabia (posing a
high temperature–based degradation). The two environments represent different
forms of erosion on critical missile components. The use of RFID and integrated
sensors allows for the complete life cycle environmental data to be used to assist
in avoiding unnecessary demilitarization of such high-value items. This example
does well to clear the Zrimsek Hurdle — it derives value in a new process through
the use of RFID-collected data. Asset services agents could be dedicated to
collecting and aggregating the environmental data for all 200 missiles. This would
guide ordnance evaluators in determining the right statistical sample of missiles to
test to determine the best demilitarization point.
Suppose a shipment contains a chemical warfare antidote, which is sensitive to
high temperatures impacting its shelf life. The pallet of antidote would have a
dedicated service monitoring (among other things) its temperature sensor and able
to notify local officers of a potential future spoilage issue. What if the temperature
sensor on the pallet was nonexistent or broken? A dedicated agent could first look
for agents of other pallets in close proximity to itself, obtain their temperature
readings, and extrapolate the same data for itself. If no such data was available, it
could search for an enterprise service that gives the internal temperature of the
distribution center the asset is located in. If the “3G RFID w/SATCOM” tag is used,
(which communicates directly to a satellite) the service could provide its GPS
coordinates to a building locator Web service, which confirms that the tag is
outdoors, then the object could check a Web service on weather.com to determine
temperature conditions. It could go on to detect that tomorrow has unusually high
temperatures predicted and send an alert with that information. This chain of
behavior may seem difficult to develop, but it is exactly what BPM tools can
provide with tools simple enough for a nonprogrammer to define in a quick
manner. If these tools can be provided to knowledge workers in a manner that
helps them do their jobs better, a network of asset management automation
solutions will emerge. Such a network will become a repository for asset-centric
knowledge.
Industrial Applications of the Asset Services Model
Consider a fan assembly for a jet engine, worth tens of thousands of dollars, at
rest on a shop floor. Consider the depth and breadth of knowledge and expertise
that contributes to the creation of that fan assembly, and how little of that
knowledge is accessible. The concept of knowledge management is the gathering
of information residing in people’s memory banks and storing that information in a
manner that can be reused as necessary. The asset is a terrific container for
knowledge about itself. Below is an abridged array of data and processes that
surround the asset. Note that the asset is not directly connected to the data and
processes.

21
Aftermarket Data
• Consumption/depreciation
• Life-cycle data collection
• Life-cycle services
Aftermarket Process
• Goods receipt/dispatch
• Warehouse management
• Repair history
• Life-cycle management
• Reduce fraud/theft/waste
• Returns process
• Supplier warranty tracking
• Product and consumables reordering
Warehouse/Distribution Processes
• Replenishment
• Right shipment to right customer
• Process visibility
• Tracking history
• Reduction of out of stock
Engineering Data
• Engineering revision
• Quality inspections
• Certifications
• Failure modes
• Drawings
Manufacturing Data
• Machining statistical data
• Image processing data
• Tooling, lotting
• Vendor/supplier
• Raw materials
• Instant messaging
• Quality inspections
• Bill of materials
• Routing/traveler
• Outsourced services
• Personnel data
• Project stock versus plant stock
Manufacturing Processes
• Electronic traveler
• Tracking “as built” configuration
• Automate line management
• Inventory management
• Configuration management
• Condition monitoring
• People and asset movement
• Maintenance repair and overhaul
• Mandate compliance
• Accounting/costing
Inbound Supplier Processes
• Goods receipt/dispatch
• Warehouse movement
• Supply chain movement
• Site security
• Product track and trace
• Inventory availability
Asset
ERP systems contain a great deal of standardized business process behavior that
governs many of the processes and data collection models listed in the above
figure. The asset, until now, has been only indirectly associated with data and
processes relevant to it via ERP (which affords little process change capabilities)
or shop floor systems (which are possibly more accommodating to process
changes but exist in nonintegrated “motes” of information). The following figure
depicts a means of connecting an asset to disparate data sources and enabling
localized customization building upon ERP standards. Note that in this diagram, an
asset is defined as a single asset, an asset class (such as a part class), or an
asset collection (such as a pallet or collection of subassemblies). An asset can be
any item, tangible (such as a machine) or virtual (such as a purchase order).

22
This diagram depicts a series of connections between a physical asset’s dedicated
service and the local process authority (LPA), ERP, external (Internet) services,
and motes of information. The implication is that the relationship between ERP
systems and the assets they govern changes. Until now assets were generalized
and all managed the same; in the future, assets can contain data and information
about the processes governing themselves and thus interact more intelligently with
ERP systems.
The benefit of BPM tools is that ERP standard processes can be programmatically
defined using a standardized definition language (BPEL). This may be
implemented at that asset services level. For example, an ERP system generates
a shop floor traveler to define process steps that occur in a work-in-process (WIP)
setting. At the same time an asset is RF tagged. That data is contained in a BPEL-
defined format and transmitted to the asset service, which immediately interprets
BPEL as data-defined workflow. Presently, the part is halfway through the WIP
process and any of the following scenarios play out:
1. The operator breaks the part. Corrective action is not defined in the ERP
system. The local process authority (the shop floor manager or foreman) directs
that an additional 5 steps be added to the workflow to restore the part. Visibility
to those 5 steps is lost by the ERP system, and thus is lost by the body of
management overseeing operations, as they rely on ERP-generated reports.
Using BPM tools on the asset service, the knowledge worker can quickly (in
less than 2 minutes) implement a routine to account for the change in workflow
and subsequent product delay. The standard routing is preserved, just with
additional steps placed within it. The routine may be reused at will and it turns
out that it is used surprisingly often (similar parts are routinely broken on that
work step). An engineering change on the part and process step is
subsequently raised. A series of instant messages between the design
engineer, process engineer, and machinist ensues, all of which are recorded to
the asset service. The solution requires iterative problem solving steps, which
are all documented in the object containers dedicated to each asset. Over time,

23
a library of custom process steps for such contingencies is implemented in that
WIP area and available for other organization units to adopt.
2. The part is placed on a truck to be shipped to another company plant. The truck
driver leaves for the weekend before making the delivery. The company has
lost visibility into where the asset is.
The plant manager creates his own workflow rule to notify the relevant parties
when an asset is loaded on a truck in one location and does not show up at its
intended location within a reasonable driving time.
3. The part is in one of four production lines which all end with the same finishing
process. The end of all four lines is stacking up with inventory ready to move.
Presently the finishing process is idle, as none of the WIP inventory has been
moved yet. A bottleneck in operations is about to occur.
The plant manager builds a rule to monitor WIP inventory levels in all pre-
bottleneck locations. Notifications are sent to the line managers and manager
of the finishing process.
4. An expensive tool used in forming the part breaks. The tool is placed in a scrap
tooling bin.
The asset service detects the tool’s location and reminds the relative worker
that the tool is still under warranty, and offers a link to the warranty claim
process.
5. A new tool, an automated torque wrench, is used in the WIP process. The new
tool requires torque-related data.
The asset service interfaces with the tool’s controller to provide torque
specifications derived from either its BPEL-based shop floor traveler download,
or the asset service calls an enterprise service which provides the data for that
part type. The ability to override the torque settings is limited to shop process
engineers.
6. The asset is finished and incorporated into an aircraft engine. Years later, the
engine is overhauled and the asset is found to be severely damaged, requiring
investigation. Some, but not all, similar assets have the same failure modes.
The original asset services used when manufacturing the failed parts are
restored from a persistent state. All of the data collected via the asset services
is made available to forensic engineers. The root cause (e.g., an improperly
calibrated tool or particular engineering revision level) is found and corrective
action taken, including a recall of other specific parts that were impacted by the
defined root cause. The manufacturer is able to limit the recall to the right parts.
Data can become a form of intellectual property that provides competitive
advantage. Further details are available in the “Aircraft Engine Aftermarket”
appendix section.
The network effect is at work. Over time, a series of routines become
commonplace, some account for process steps that are ill-defined in the ERP
system, some reflect specific workflow notification tasks which help management
detect issues, some reflect rules such as invalid part combinations, etc. Beyond
allowing adaptive process changes, this proposed model offers a platform for
integrating disparate data sources that may or may not be relevant to ERP.
Benefits of an Asset Services Infrastructure:
• In-memory processing: continuous data streams may be monitored by an in-
memory mechanism with a robust set of tools to aggregate data and/or apply

24
business logic to data streams, triggering events such as ERP transactions or
employee notifications.
• Data and process change management: because the process is tied to an
independent object, legacy processes may be allowed, enabling perfect
flexibility in upgrading processes. New data sources may be tied to the asset.
• The asset service allows robust asset management even when ERP systems
are unavailable due to WAN outages or maintenance outages.
• Inheritance of standard behavior: core ERP-driven standards are implemented;
the local process authority cannot override core process requirements.
• Scalable platform: the object may be extended to provide and consume an
ever- increasing amount of enterprise and Web services.
• Object interaction: As demonstrated in the Age of Empires, objects are able to
execute arm’s-length transactions with one another.
• Ability to rapidly prototype: new processes may be implemented and changed
while in process one asset at a time.
• Platform for “lean” management: the platform enables an event-driven
environment, dovetailing with the single-piece flow concept.
• Platform to enable computing “intelligence” in deriving inferences about real-
world events: offers significant functionality for system security (see the CEP
entry in the appendix for more details).
Recommendations
While the asset services model is not a mature product available commercially
now, an IT decision maker can take steps toward enabling the network effect by
looking for the following demonstrated principals in software purchase or
development decisions:
1. To harness the network effect is to enable solution participants to add value.
This is accomplished by providing simple, light development toolsets to
knowledge workers. This requires defining the knowledge worker population
and giving that group a voice in the software selection process. Note that
process authority distribution will vary widely; for example, Wal-Mart has a
centralized process authority (processes are not different for each store or
distribution center), whereas GE has a very distributed process authority (with
the strong emphasis on Six Sigma, which drives continuous improvement down
to the front line managers). A Wal-Mart solution will likely be customized by a
centralized competency center, which will derive and implement global best
practices. The GE solution will likely be used by varied, distributed process
owners, and the resulting applications will be unique to each division; however,
the tools used to build each unique application will utilize a standardized
toolset. The nature of the toolset will be determined by the technical expertise
of the body of knowledge workers. Therefore, when comparing RFID candidate
systems, look for the ease of solution customization in the hands of knowledge
workers.
2. Scalability is a critical component to any RFID system. Solutions that utilize
toolsets need to fit into a “hardening” path: a well-thought out plan for scaling
the solution several orders of magnitude. This includes the hardware
infrastructure as well as the capability of the solution to foster future growth,
such as incorporating the ability to provide and consume Web and enterprise
services. A key to enabling scalable future growth is the solution’s compliance
with established computing standards such as Web and enterprise services,

25
BPEL compliance, and EPC global standards. Multi-variable solutions will
require new kinds of enabling technologies such as In-Memory databases.4, 5
3. Solutions should mirror reality as much as possible. An RFID solution ideally
will offer a means of describing the relationships between real-world items. This
will most likely leverage an ontology which will describe such relationships.
RFID adoption should be included in strategic planning, as an overall direction
towards an asset services model will not happen otherwise. RFID systems are
known to accelerate the speed of business process change, prompting
organizational changes. The procedures used to facilitate business process
changes may need to be redefined. The organizational structure may need to be
altered. A competency center may be needed to handle external infrastructure
integration, toolset hardening paths, and support for the local process authority.
Strategic planning needs to extend beyond supply chain solutions and consider the
value of RFID in non-product assets or as a means of delivering value-adding
services to the product offering. The Medtronic Corporation makes implantable
healthcare devices such as pacemakers. In the 1990s, Medtronic built a strategic
vision that planned a service offering by 2010, which would allow implanted
devices to communicate vital statistics via RF signal to a patient’s home-based
reader, which would transmit this data over the Internet to an online healthcare
center. Here, RF technology is leveraged with the external infrastructure of the
Internet to deliver a strong product differentiator. Medtronic can leverage the
standards and network available over the Internet while preserving their value
proposition with a proprietary protocol in the device-to-reader exchange. The “own
versus influence” model offers some guidance when making strategic decisions
surrounding the patching of emerging technologies. This model suggests that
technologies with a potential for proprietary technical advantage be developed,
procured, or acquired (see the appendix for the model diagram).
RFID usage breaks into two categories. The first category is the use of cheap
(costing pennies), disposable, passive RFID tags running serial process
optimization in an open- loop environment. This category will be used to optimize
existing business processes. It will utilize various external infrastructures such as
the EPC global network and SCM hubs (such as SAP’s in-development SCM
hubs, or vertical search engine communities).
The second category will utilize more costly RFID tags, (ranging from basic active
tags to the 3G RFID w/SATCOM tag, which is globally visible) and will utilize an
ontology of defined relationships, leveraging the network effect and possibly
implementing CEP for system security. In this category, RFID is used to visualize
the status of multiple groups of assets. Integrating systems like IMMACCS with 3G
RFID w/SATCOM in military applications is likely to be the first widely adopted
implementation of this kind of RFID usage.
As the market leader of ERP solutions, SAP is leading the way in building context-
sensitive, industry-specific solutions built on reusable services. SAP is providing
4
Some instant messaging solutions offer the ability to directly view server memory over the
network. This is a pull-based means of viewing data (and is heavy on network bandwidth);
however, this offers a highly scalable means of dedicating any number of network servers to
a single stream of data.
5
Multi-variable data sets offer more potential value as the permutations of variable data
combinations offers exponentially greater data specificity about the real world. For example,
a single data set tracks the location of WIP assets. Another data set tracks tooling
locations. When combined, (making a multi-variable data set) the system can deduce which
tools were used on which assets, offering functionality not available when the two data sets
were separated.

26
avenues for joint “co-innovation” amongst solution providers through a common
business language. This business language connects disparate infrastructure
products and software applications in its “partner ecosystem” space. According to
SAP, the speed of innovation is increasing.xxi
These developments intersect with the RFID data collection world to enable a
fundamental change in the relationship between assets and IT systems. Assets will
become contributing nodes in a network of solutions that offer new services to
consumers and new efficiencies in asset management. As with the “horseless
carriage,” RFID will grow out of the conception of “wireless bar coding” as it is
adopted in business, military, and society.

27
Appendix
The Aircraft Engine Aftermarket
Unique identification allows one to ask and answer valuable questions from a
maintenance and repair operations perspective. For example, consider a jet
engine hollow fan blade disassembled from an engine being overhauled. That
blade has a unique identifier (called a UID, which is a two-dimensional bar code
etching onto the part). Management can know the life-cycle history of that part,
plus on-wing usage information for engine diagnostic systems. Inexpensive image
collection and processing software can automate the identification of the “failure
modes” exhibited by that blade (the failure modes being visible signs of wear and
tear to the part). Here is an abridged list of available information regarding that
particular blade:
• Aircraft engine engineering build: an aircraft engine will have a specific
engineering revision build, much like how an operating system or application
has a service pack level. Differing levels will impact component wear or
depreciation differently
• Life-cycle data such as on-wing vibration, excess temperature, even plane
operating locations (e.g., a plane in the Middle East will incur more sand-
oriented damage)
• Value chain data such as raw material supplier (even which hole in the ground
the oar came from), engineering revision level, lot data, assemblers,
machinists who built the part (and their relative certifications), tools used,
calibration of the tools used, etc.
• Other parts in the engine, with their relevant value chain and life-cycle data
• Historical data on failure modes of all blades
Here is a dramatic example of the vast amounts of data that could be integrated to
better predict part depreciation, provide feedback to engineering on failure modes
and what usage scenarios caused them, and better predict spare parts
requirements, allowing for reductions in inventory. This demonstrates the concept
of data as a competitive advantage. The manufacturing data generated by a
company cannot be duplicated by competitors. Any value derived from that data
may not be duplicated. This can make data a form of intellectual property.
Pratt & Whitney announced in February 2006 that it would begin making parts for
GE engines. A jet engine manufacturer sells its engines at cost, and makes
several times that amount in the aftermarket which spans from 20 to 30 years
following the sale. This move by Pratt & Whitney substantially impacts GE’s
financial models, and could have a ripple effect across the aerospace industry.
This demonstrates the potential business impact of aftermarket ownership, as well
as the concept that your manufacturing data could constitute a competitive
advantage. Data can become a form of intellectual property; no other company can
duplicate it.

28
Details on SAP’s “Auto ID Infrastructure” (Aii) Product
The Aii module is designed for the first scenario shown above (as indicated by the
“1” and the green connecting lines). The Aii module reads device controller-
provided data serially (one record at a time) and applies rules, or filters, to the data
to determine what activities or function modules are executed. This is ideal for all
gate-based activities. The Aii module interfaces via XI (SAP’s EAI toolset) to R/3
and other SAP components. Aii also has its own Business Warehouse stack. Aii
can trigger automated transactions in R/3, such as a goods receipt or shipping
transactions. If customized, other transactions may be automated, such as a
goods movement in inventory. Aii has a mobile device component, so SAP-
delivered mobile applications will enjoy more rich functionality as the Aii product is
continually invested in by SAP. Aii can also send data to the Event Manager, which
is an SAP component that provides process track and trace functionality.
A multivariable processing solution (e.g., an in-memory database) may be placed
in front of Aii (see the “Localized Solution” box above). Such a solution could glean
events from RFID multivariable data streams and pass that to Aii for processing
(marked as “2”). If the Aii module does not need to be used, the solution could
connect to the XI (EAI hub – marked as “3”) data processing hub, which will route
messages asynchronously to targets, such as the event manager. Lastly, if a
highly customized R/3 application requires synchronous communication with the
RFID device controller layer, the line marked as “4” may be drawn by implementing
a SAP JCO solution which utilizes a direct “RFC” connection. Another option (not
displayed as it is currently not validated by SAP) is to connect RFID asset location
information directly to the cache on a custom table located on an R/3 application
server. This would be used when frequent random R/3 calls are issued looking for
specific asset location data.
The “Heavy Solution”, (see diagram below) displays the cost of implementing Aii
for a heavily customized R/3 application. In this case the Aii software is purchased;
however, the standard Aii functionality does not fit with the existing custom packing
solution. The result is that Aii is used as an EPC generator and device interface.
The solution is heavy in system resource consumption, infrastructure component
requirements, and ongoing maintenance. If a data structure is changed in R/3, the

29
change could require up to 6 downstream data structure modifications (steps 1, 5,
7, 11, 15, 18).
The “Light Solution” is an alternative that involves fewer moving parts, faster
execution time, less infrastructure resource consumption, and more code
reusability. The connection between steps 3 and 4 of the light solution is a point to
point connection that is authenticated once then kept open as long as both
components are live. While XI offers the ability to add future subscribers to the
data flowing through it, it also validates each message’s header login data. In the
case above, it is very hard to imagine a situation where another network entity
would reasonably wish to receive the data being transmitted to the device
controller.
A benefit of RFC-based communication with SAP is that a session is established
and persists beyond each message. The Exchange Infrastructure product requires
that each message packet be authenticated, adding data, overhead, and
functionality not needed in this example to the data transfer process. This leads to
the following rule of thumb:
Where very high volume data transfers occur on an ad hoc basis between a
trusted on-network source and R/3, and where such data being transferred is
highly unlikely to be of use to any other potential applications, such a connection
may utilize the RFC protocol, bypassing XI. There are more variables to take into
consideration when choosing the best solution (consult your local CSC technical
representative).
The value of Aii is primarily in its extension of SAP business processes and the
development platform it offers. For this reason, it appears that the best
implementation of Aii is as a single, integrated solution co-located with its target
R/3 server. The architecture can scale by adding device controllers (with
expandable data filtering functionality) placed in various shop floor locations,
matched by Aii server resource additions which allow increased work processes
and communication channels.
Complex Event Processing (CEP)
CEP is a means of building causality models which can interpret multiple sources
of data to gather inferences about real world events. CEP does this by defining
event causality within a system. For example, a tagged asset is moved from a WIP
location to a finished goods inventory location. The RFID access points transmit
the data to a location appliance, which is interrogated by an in-memory database.
That database applies established business rules to detect the WIP completion
event and triggers an event by sending a message to a SAP Aii module. The SAP
Aii module receives a single data packet, applies a rule, which triggers an activity
and processes the event by sending a message to SAP R/3, which performs a few
validation checks and then writes to the database. This is graphically depicted
(partially) in the below figure.

30
This figure represents streams of data from RFID access points being filtered for
events, which cascade down to a layer of activity (or persistence). The small green
square at the persistence layer represents the WIP to Finished Goods Inventory
event. That event (referred to as the “causal event”) triggers an ERP application
server function to perform 3 sub-functions, the last of which writes data to the
database server. The hierarchy of activities caused by the one causal event is
denoted on the right hand side of the figure. Note that systems work processes
and network transmissions are included in the hierarchy.
Among other things, the CEP model creates a system monitoring mechanism that
anticipates what the system load should be, given the inventory of current causal
events. It can then check what the system load actually is. A discrepancy between
what should be and what system load actually is may indicate a security breach.
For example, an unexplained spike in network activity may indicate a denial of
service attack. An unexplained set of server work processes may indicate the
presence of a software virus. The basic tools that CEP uses involve the tracking of
causality and timing of events in a system landscape.
A major concern about RFID is security.6
A CEP model may be applied to help
secure a system from unauthorized access, and may be used to compile more
variables about real world events to gauge suspicious behavior.
6
Though security is out of scope of the grant work, (it is addressed in a 2005 LEF Grant) the
recommendation by the grantee is to avoid storing sensitive data on an RF tag, instead, store that data
on a network and use the tag serial number as a data access mechanism. If the secure data is required
in areas without network connectivity, consider a UID label (which would require a direct physical read
by the user and a camera-based reader). Other options involve using more advanced RF tags, for
example if concerned about unauthorized RF reads of items in transit on a truck, the RF tags can be
designed to transmit when not in motion, or only when not in motion for a certain amount of time. Lastly,
the EPC Generation 2 standard allows for tag readers to supply encrypted passwords which active tags
validate before responding.

31
CEP defines the Event Processing Language (EPL) which can be used to answer
in-memory database questions (such as “tell me when these two events happen
within 30 seconds”) as well as interpret data streams to infer real world events
(e.g. the asset is on the lathe, the lathe is on, the worker is in front of the lathe,
therefore the person is working on the asset).
The Own Versus Influence Model
i
Stakutis, Chris, Webster, John. “Inescapable Data: Harnessing the Power of
Convergence.” IBM Press, 2006, page 1, Chapter 8.
ii
Stakutis, Chris, Webster, John. “Inescapable Data: Harnessing the Power of Convergence.”
IBM Press, 2006, page 2, Chapter 8.
iii
Collins, Jonathan. “RFID’s Impact at Wal-Mart Greater Than Expected.”
http://www.rifdjournal.com/article/articlepring/2314/-1/1 read 5/5/2006.
iv
Collins, Jonathan. “RFID’s Impact at Wal-Mart Greater Than Expected.”
http://www.rifdjournal.com/article/articlepring/2314/-1/1 read 5/5/2006.
v
ComputerWeekly.com, “RFID Revolution Years Away, says Gartner,” July 29, 2005.
http://www.computerweekly.com/articles/article.asp?liArticleID=135349&liFlavourID=1&sp=1
#
vi
Collins, Jonathan. “DoD Quantifies Payback from RFID.”
http://www.rfidjournal.com/article/articlepring/2323/-1/1/ read 5/5/2006
vii
Collins, Jonathan. “DoD Quantifies Payback from RFID.”
http://www.rfidjournal.com/article/articlepring/2323/-1/1/ read 5/5/2006.
viii
“New Portable Deployment Kit uses Iridium Satellite Links to Track Military Supplies in
Area of Operations,” Iridium Press. http://www.iridium.com/corp/iri_corp-
news.asp?newsid=163 read 12/5/2005.
ix
Collins, Jonathan. “DoD Tries Tags That Phone Home.”
http://www.rfidjournal.com/article/articleview/1458/1/65/ read 12/5/2005.
x
Fee, Jeffrey D. and Schmack, Alan. “Improving RFID Technology.”
http://www.globalsecurity.org/military/library/report/2005/050200-rfid.htm read 12/5/2005.
xi
“Architecting the Internet of Things,” http://autoid.mit.edu/cs/video.aspx
xii
http://www.scripting.com/disruption/mail.html

32
xiii
http://www.internetnews.com/ent-news/article.php/3562881
xiv
Greene, Jay. “Microsoft’s Strange Spending Splurge.”
http://webdesign.ittoolbox.com/news/display.asp?i=142080&t=7 read 4/28/2006
xv
http://en.wikipedia.org/wiki/Groupthink
xvi
“Q&A with James Surowiecki,” Random House.
http://www.randomhouse.com/features/wisdomofcrowds/Q&A.html
xvii
Luckham, David. “The Power Of Events,” Pearson Education, Inc. 2002, page 21.
xviii
Pohl, Jens; Wood, Anthony A. (Col. USMC, Ret.); and Pohl, Kym Jason. “IMMACCS: An
Experimental Multi-Agent C2 System.” http://www.cadrc.calpoly.edu/pdf/immaccs_multi.pdf
xix
Pohl, Jens. “Intelligent Software Systems: Do We Need Them?”
http://lef.csc.com/technologyprograms/
xx
http://en.wikipedia.org/wiki/Age_of_empires
xxi
Agassi, Shai, SAP Executive Board Member. “Leading the Web Services Wave.” (Blog
Site) http://www.sap.com/community/int/blog/ShowBlog.epx?PostID=353432 read
5/22/2006.

RFID and Agent-Based Computing

More Related Content

Similar to RFID and Agent-Based Computing (20)

RFID and Agent-Based Computing