Nowadays production and distribution processes are controlled to a large extent by information processes. Interconnecting the ow of material and information promises to increase e ciency and quality of business activities while reducing costs. Production planing and control systems and enterprise resource planning systems can optimize processes within companies. Supply chain management comprises the whole process of planning, implementing, and controlling the operations within a supply chain. The overall goal is to satisfy customer requirements as e ciently as possible while saving time and resources on the other hand. Controlling all logistic processes from raw materials suppliers to manufacturers and retailers to the consumer needs transparency of the process steps. Managing the whole product life cycle from point-of-origin to the point-of-consumption and disposal or recycling needs accurate information about products in each process state and at the appropriate location. Automatic Identi cation and Data Capture (AIDC) or Automatic Identi cation (AutoID) comprise techniques to automatically identify objects, to collect associated product data, and to propagate that data to back-end software applications like Enterprise Resource Planning systems. Typical automatic identi cation technologies are bar codes, Radio Frequency Identi cation (RFID) and smart cards. These techniques identify assigned properties while biometrics, optical character and voice recognition utilize natural properties. RFID has its origins more than fty years ago. But due to recent developments this technique is now available with higher quality and more functionality at lower costs. Especially in logistics a large-scale commercial application is expected. In [Hen09] an worldwide market increase of 25% p.a. is expected.
RFID technology comprises several aspects: infrastructure and architecture, hardware like tags or transponders and readers, integration on the physical layer up to IT system integration: SCM, ERP, MES and warehouse management systems.
System architects, technology integrators, process designers and engineers in charge of implementation and system integration face several challenges. On the physical level rst of all "material things" have to be identi ed. Then the captured data can be processed, ltered and forwarded to applications which perform planning and controlling tasks. Questions concerning appropriate data formats, e cient data transfer, data integration, lookup mechanisms, and others arise in AutoID environments. This pattern collection introduces patterns dealing with aspects of those problems.
This collection is a continuation and extending of a work starting with "A Pattern Language for Process Optimization with Smart Object Identi cation\ at EuroPLoP 2005 [Bie05] and \Patterns for Unique Product Identi cation" [Bie08].
A decentralized data storage has bene ts in applications where access to a network infrastructure is temporarily or in principle not available. In such scenarios Data Accompanying Products is an appropriate solution. Storing product data decentralized attached to the product and at the same time centralized on servers can easily result in data inconsistencies. Synchronised Data Location explains how to cope with that problem.
Production automation and logistic processes
Need for centralized product data management
Need for decentralized
data access Identifiers Point To Data
Data Accompanying Products
Cope with large amount of raw data
Avoid inconsistency when storing data centralized
and decentralized Context Identifiers Point To Data is necessary to enable seamless access to product data within a single company’s internal production processes as well as in case of supply chains where several companies are involved.
How to give access to detailed product data in distributed production and logistic networks while only an identi er is attached to the product? In production and distribution processes several partners needs to gain access to detailed product data. If the product data representation is attached to the product itself the describing information can only be obtained when the product is present.
Product data not only has to be accessible but also has to be updated and maintained during the whole product life cycle from raw material suppliers to customers and nally to disposal or recycling. Optical data representations are printed once and afterwards unchangeable. Modi cations can only be done by replacing an old label or product accompanying description by a newer one.
Machine readable data has to be stored in such a way that it is assigned to the describing product. Typically a label is xed on a product or it's packaging. Alternatively the product itself is marked (Direct Part Marking) e.g. by means of laser engraving. Relatively cheap RFID transponders store about a hundred Bits (EPC speci es a 96 Bit encoding) up to one or two kBits. The amount of data is restricted.
An unique identi er allows distinction of product types or even individuals. But this is not su cient if a detailed product description is necessary e.g. for customers or due to documentation reasons.
Increasing complexity of several companies spanning production processes and deep supply chains raise the demand for product data accessible beyond company internal and isolated IT systems like Production Planing Systems (PPS) and Enterprise Resource Planning (ERP) Systems.
Establish a network-based infrastructure that consists of components providing services for registering resources with an associated data lookup service which enables storage and query of data along with their identi ers (keys). The services may be be folded into one single piece of software or can be distributed over the Internet.
In real world physical objects populate the environment. Products or transport units are a speci c type of physical objects as results of planed and controlled production processes or as part of such processes or as elements of transportation and storage. Information technology supports controlling of technical or administration processes. Processing and storing data about the real world needs abstract model of reality. Objects of the real world are abstracted in the information systems through the use of symbols (names, labels).
Abstraction is necessary because the real world cannot be represented in each detail, but relevant sections of reality have to be chosen. Data describing properties of the physical objects may be extensive. Often several parts of organisations or individuals especially customers and users are interested in the data belonging to a product or transport unit, even if the physical object itself is not present.
Rationale Accessibility to product data can be gained by saving and processing the data in decentralized databases. If these databases can be accessed over the Internet most of the involved partners will nd the data easily. An identi er has to be assigned to products or transport units and to be carried along with them which works as a "pointer" to it's belonging data. In this sense the identi er externalises the identi er that is necessary internally in databases to mark each entry uniquely. RFID is seen as an enabler so that things of the real world have a representation in the \virtual" world of the Internet [KBM+00], [Bie05].
Product data can be accessed independently of the products presence. Only the identi cation number has to be represented or stored. Hence printed labels or inexpensive passive RFID tags are su cient.
Generation, representation and assignment of identi ers for products or transport units is well standardized. The standards speci ed for reading devices, middleware, and data manipulation grant compatibility and ensure a long-term use of AIDC technologies like barcode, two-dimensional symbologies and in an increasingly regarding RFID.
Data accessibility is highly dependent on the availability of network infrastructure. Centralized data storage and processing has to be reachable from every decision point within the production or logistic processes. Failure of the centralized IT system or network components result in a stop of the process. Redundancy can reduce the risk of system standstills, but raises costs.
Network tra c increases with the number of items which are stored and have to been searchable. The e ort of resolving identi ers, searching the data and delivering it to the requester grows proportional to the amount of item records. Centralized data processing induces demands for data access regulations. Security issues may be raised by producers and customers as well: To avoid business espionage producers need to hide product details from their competitors. On the other hand customers have concerns that their consumer's behaviour may be recorded and then misused by unauthorized parties.
A well known use are cash registers at checkout points in supermarkets which scan the bar code on products to obtain data like price or weight. The data can be stored decentralized at the cash register itself or is stored on a centralized system for the reason of easier updates.
In logistics nearly each returnable transport item like pallets or containers are assigned a global shipping item number such that the transport units can be tracked. Interoperability and ease of use need standardisation. Global Standard 1 (GS1) has de ned a scheme for globally applicable identi ers for trade items (products and services), the Global Trade Item Number GTIN. A whole family of data structure are speci ed for di erent application areas. In case of consumer products the EAN13 bar code is very popular especially in Europe [GS1]. The symbol encodes 13 numbers and is divided into four parts:
System code: two or three digits representing the country where the manufacturer is registered. To cover the International Standard Book Number ISBN and the International Standard Serial Number ISSN the starting three numerals 978 resp. 979 are used.
Manufacturer code: four or ve digits such that system code and manufacturer code are 7 numerals in sum.
Product code: ve digits given by the manufacturer (normally the serial number).
Check digit: one digit for the check sum. The EAN identi er can be scanned to look up product information especially a description and the product price in a database. As a service for consumers GS1 provides a web site on the Internet (http://directory.gs1.org/gtin/search) where the product code owner and item information can be searched for. Several courier-, express- and parcel (CEP) service provider o er customers freight tracking services. In each relevant step in the logistic chain from sender to handler and nally the the recipient is documented by capturing identi cation data and associating it to the ful lled transaction. As a service customer can look up information about their shipments. Also Known As: Data-on-Tag
Application of Data Accompanying Products enables process automation where data about physical objects like products or pallets is necessary.
How to provide machines access to product data without presence of a network infrastructure?
A centralized data storage allows data access from several distributed decision points in production and logistic processes. If the necessary network is not available or if the centralized data storage and processing system fails the whole process stops.
Identi ers attached to physical objects in the real world operate as pointers to resource in the \virtual " world. If the only information carried along with a physical object is the identi er then no detailed data can be obtained between points with no network access.
But in case of unplanned events more detailed data like product contents may be necessary which cannot be obtained abroad from the normal process paths.
Network tra c and lookup times increase with the amount of items for which detailed data is requested.
Use a machine readable representation or storage technique carried along with the product itself such that product related data accompanies the physical object.
In order to attach data to a physical object in a machine readable manner an appropriate representation is necessary which is capable to store data and allows at least a reading access.
Read only techniques can be distinguished due to the used storage media and communication method. The most common solutions are:
Optical representation with transfer in the frequency range of visual light. Data is encoded as linear or two dimensional arranged geometrical patterns like rectangles, squares or circles. The encoding is based on the shape of the geometrical element and/or it's location.
Typically the patterns are printed on adhesive labels which are then attached to the physical object. Alternatively a direct part marking by means of Laser engraving or other surface processing is applied to prevent removing of the data. Data storage in digital memories and communication devices on chips. Data transfer is performed via radio frequencies. On principle, storage capacities of digital memories can be very large but data transfer then is time consuming. While printed or directly marked symbologies are immutable except for replacing the label or modifying the marked surface, digital memories provide the opportunity of not only reading from but also writing to the chip.
Rationale Product data accompanying the the product itself grant access to information without a network infrastructure and a centralized data processing [DMS07].
Attaching human readable information to products is well established for decades. Examples are product sheets or instructions for use which are printed separately and packaged together with the product. Shorter instructions and descriptions are printed on labels and xed on the product. Textual representations are not obsolete and to be replaced by machine readable data. In fact machine readable information augments the \audience" such that devices can get read the data which is then can be exploit to automatically control process steps or is displayed to humans. Attachment of data beyond simple identi ers raise the demand for higher storage or representation capacities. In case of optical identi cation techniques data capacity was stepwise enhanced in the evolution from linear codes { bar codes { to stacked codes and then to two-dimensional symbologies. A widely used two-dimensional symbology is the Data Matrix code which can contain up to 3116 numeric or 2335 alpha-numeric characters.
The digital memory integrated in RFID transponders feature a capacity of typically about 100 Bit su cient for identi ers up to several kBit. It is guessed that Moore's Law just begins regarding RFID transponders.
Access to product data can be accessed independently of the products presence. In case of re-writeable data storage labels like RFID transponders relevant data can be collected on the path a product is undergoing.
Object related data allows managing of processes without a centralized controlling system.
Data is available without an IT infrastructure and without the need for a centralized storage and processing system.
Detailed data is only accessible when the described physical object is present. Data storage capacities limit the amount of information. The level of details hence is restricted and not all relevant data might be kept.
Siemens Amberg produces transformers which can be individually customized. Figure 4 shows a transport unit used in the transformer production process at Siemens Amberg. Individual con guration data and necessary production steps are stored in the transponder at the beginning of the production line. After that the transport unit is directed through the production and the speci ed steps are performed automatically without a centralized production execution system. Further applications of RFID in process automation are Ford's manufacturing process at the Essex Engine Plant in Windsor, Ontario, and the aircraft maintenance at Boeing and Airbus [ZGYP03], [ZGP03].
In the United States the Food and Drug Administration (FDA) forces the introduction of a so called ePedigree for pharmaceutical products. In California by law prescription drugs have to be accompanied by an electronic pedigree. In the US RFID is the preferred tracking technology. Due to problems of wide spread application of RFID transponders, hybrid pedigree systems incorporating paper and automatic identi cation technologies like RFID or optical codes are in use as shown in gure 5.
Synchronized Data Location is needed if data is alterable at both the data carrier attached to the product and in centralized storage.
How to avoid data inconsistencies if product data accompanying a physical object itself and the mirrored data on centralized servers are changed independently?
A centralized data storage allows data access from several distributed decision points in production and logistic processes. But if the infrastructure is not available or if the centralized data storage and processing system fails the whole process stops.
Carrying product data along with the product itself provides access to that information independently from a network infrastructure { stand-alone reader devices are su cient. But due to the decentralized storage technique necessary information is not permanently and accurately available for tasks on a system wide level like production planning and controlling.
Integration of all reader devices in a network infrastructure can help to forward product related data to centralized processing services. In this case data is pushed forward each time a physical reading process happens. On the level of businesslogic applications the need for information about the status of physical components is triggered by software execution. Hence, data generation and information processing are not synchronized. But the ability to react immediately to physical events needs accurate process data.
Storing product data in both ways { attached to the product itself and on centralized servers { often results in data inconsistency. E.g. updating the product data attached to the product at each process steps provides detailed information about a product’s history at any time and anywhere just having a mobile reader device at hand. But that information tends to di er increasingly from the centralized stored data.
Synchronize product data carried along with the product and the centralized stored data by means of synchronization rules.
Such rules de ne actors and execution details of data interchange processes. Important actors are data producer and consumers who are assigned updating and writing access rights. Synchronisation with the centralized system is mostly based on data versioning and event-based rules.
In automation and logistics planning and controlling applications need data related to products and process progress. Modern AutoID technologies support those applications. The overall aim is e ciency based process transparency: to know as much as needed about state and location of products or transport units. Traditional identi cation data like Barcode labels or imprints are unchangeable. Hence no data inconsistency can occur (expect in case of damage or replacement of the ID). Radio-frequency identi cation allows storage of data far beyond simple identi cation numbers. Data Accompanying Products explains bene ts of decentralized data storage. The resulting risk of inconsistent contents between separately located data not only arises in the case of modern automated identi cation and data capturing technologies: it is a general problem that occurs in backup scenarios and distributed systems. Hence AutoID infrastructures can bene t from solutions established in other IT application areas.
Centralized stored data can more accurate represent the actual state of physical things within the monitored process.
Decision and controlling systems can react more promptly to occurring changes within processes.
In case of writeable storages attached to products like writeable RFID transponders central data updates can be downloaded and deployed to the de-centralized memories on-site.
Data consistency is determined by the completeness of the synchronization rules. Unaccounted exceptions may still result in data inconsistencies.
If the necessary network infrastructure is not available no synchronisation can be performed. Data accuracy and hence quality on the centralized storage is dependent on refresh periods.
Based on [Sch02] researchers at University of Bremen / BIBA developed the concept of \Data Contracts" for AutoID applications. Data Contract specify how to update and match data between locally and centralized stored data during a product’s life cycle.
Updating and synchronisation of product data during its whole life cycle is the aim of Product Lifecycle Management PLM. From production to distribution, reselling, consumption and recycling all generated and necessary data about the products of a company have to be managed consistently and have to be provided to controlling and planning systems [AG06]. Automatic identi cation techniques contribute to PLM on the narrow layer of plant automation and production or distribution logistics.
Business Event Managers bridge the gap between streams of raw sensor data and applications on the business-logic level.
How to perform process controlling e ciently while on-site reading devices can only deliver raw sensor data which might even be redundant? Automatic identi cation techniques sense physical properties of things like given properties or attached identi ers. Controlling of production and logistic processes is based on logical rules which determine control ows.
Auto-ID technologies will enable businesses to move from linear and manual supply chain planning and execution to an event-driven, adaptive supply network. But devices like bar code scanner and RFID readers can generate multiple readings of the same physical object.
The occurrence of a reading event is not necessarily a business relevant event. Propagation of each single reading event to the planning system may result in high network tra c.
Augment the automatic identi cation and data capturing infrastructure by additional on-site software that performs a preprocessing of the raw data and propagates business relevant events to the process controlling applications. Such software components act as business event managers. Business event managers have to react on real-time events and information, to propagate alerts, and to provide services for essential reading and - if applicable { writing of information to other information systems. As shop- oor near components they have to provide interfaces for reader devices like scanners or RFID readers and other devices like sensors. To be easily integrable those components need standardised network interfaces. To be accessible from the business-logic level business event managers have to support standard communication protocols like Simple Object Access Protocol (SOAP).
Possible eld of applications raise with abilities of general-purpose event routing, collating, and ltering.
Rationale Business event extractors and managers hide hardware speci c interfaces and extensive streams of raw data. Furthermore, they can accomplish hardware con guration tasks and maintenance related services.
From the physical layer consisting of hardware devices like scanners and readers only physical signals can be obtained. Devices on this layer have to been managed e.g. by providing appropriate software like drivers.
Signals have to be transformed into data due to speci c protocols and grammars in order to be transferred to the next layer. So at least two tasks have to be done on the hardware layer: device management and data extraction and communication [GH06].
The top layer of business process management can be seen as the opposite of the hardware related layer on the bottom. Back-end systems like Enterprise Resource Planning (EAI) solutions, Manufacturing Execution Systems (MES) or Supply Chain Management (SCM) systems need to know the status of the underlying processes from a business perspective. Hence business process relevant events or alerts are necessary for decision making and then commands are sent down to the active process elements to control progress. So extraction of business relevant events and command propagation are the tasks to link the business controlling layer and the layer of process near sensors like readers and actors.
Assignment of speci c tasks and functionality to specialized components supports exibility and re-usability. Especially in case of multi-tier architectures the necessary processing power is at the point where the relevant data has to be captured and preprocessed.
Liabilities The increasing amount of components and communication overhead are time costly and decrease performance. Since more processing power is allocated at distributed hardware components the installation costs of the whole infrastructure grow.
In case of complex production processes or supply chains several distributed and heterogeneous components have to be integrated. This is the task of middleware which is referred to as Enterprise Application Integration (EAI). In between of those layers several functionality is necessary:
Preprocessing of the raw data received from devices, data ltering and aggregation, context-based event extraction, propagation of business relevant events and connectivity components for communication.
This intermediate layer, which might be subdivided, is known as \edgeware". On the edgeware layer several tasks have to be done. Devices are distributed at decision relevant points of the process. Often inhomogeneous hardware components are in use. Software components specialized in those di erent tasks and adapted for the di erent hardware devices are the constituent parts of the edgeware layer. Edgeware software is a good example where to apply the Layers [BMR+96] and Pipes and Filters [BMR+96] architectural patterns.
Distinct parts work di erent levels and speci c protocols can be used to access the services contained at each layer.
In case of a single tier architectural approach all software functionality for data and device management is integrated as a single software component with a layered internal structure. Figure 6 illustrates a single-tier edgeware architecture.
Figure 6 { Single-tier Edgeware Architecture In production plant automation and in logistic applications reading devices are distributed. In such a setting a multi-tier architecture is more appropriate, as shown in gure 7. In this architecture an additional service bus separates command interpretation and event extraction from hardware speci c functionality. The latter consists of command transformation into device speci c formats as well as data ltering and aggregation. In [Lea04] several industrial applications of one-tier and multi-tier architectures are introduced.
Figure 7 { Multi-tier Edgeware Architecture GS1 aims on a standardisation of integration architectures. That standardisation de nes the The Savant Architecture [LNE05], [CTAO03] where interface components on the hardware near layer perform simple data processing in order to extract Application Level Events from raw sensor data.
Figure 8 { Event Manager Software Components in RFID Middleware
Left: Sun's RFID Middleware (Source: Sun Microsystems)
Right: Microsoft's BizTalk Architecture (Source: Microsoft) The architectural approach of extracting business events is implemented in middleware platforms like Microsoft BizTalk [Sch06] or Sun’s RFID middleware [Sun05]. In both architectures specialized software components (\Event Managers") gather information from the RFID Readers, lter the information, and provide information as messages to systems on the business layer like ERP systems. Figure 8 illustrate how the Event Manager and Information Server t into a network-based AutoID architecture.
I’d like to thank my EuroPLoP 2008 shepherd Christian Kohls for his helpful feedback and inspiring suggestions. [AG06]
Abramovici, M. and M. Ghoffrani: PLM - ein Thema auch fur KMUs. Digital Engineering, 2006.
Bienhaus, Diethelm: A Pattern Language for the Network of Things. In Proceedings of 10th European Conference on Pattern Languages of Programs (EuroPlop 2005), Irsee, Germany, 2005.
Bienhaus, Diethelm: Patterns for Unique Product Identi cation. In Proceedings of 12th European Conference on Pattern Languages of Programs (EuroPlop 2007), (to be published) 2008. [BMR+96] Buschmann, F., R. Meunier, H. Rohnert, P. Sommerlad and M. Stal: Pattern-Oriented Software Architecture: A System of Patterns. Wiley, New York, 1996. [CTAO03] Clark, Sean, Ken Traub, Dipan Anarkat and Ted Osinski: Auto-ID Savant Speci cation 1.0, 2003. http://www.nepc.gs1.org.sg/epcglobal/stdsdocs/WD_ savant_1-0_20030911.doc (17.02.2009).
Diekmann, Thomas, Adam Melski and Matthias Schumann: Data-onNetwork vs. Data-on-Tag: Managing Data in Complex RFID Environments. In HICSS '07: Proceedings of the 40th Annual Hawaii International Conference on System Sciences, page 224a, Washington, DC, USA, 2007. IEEE Computer Society. Gillert, Frank and Wolf-Rudiger Hansen: RFID fur die Optimierung von Geschaaftsprozessen: Prozess-Strukturen, IT-Architekturen, RFID-Infrastruktur. Hanser, Munchen, 2006.
GS1: EAN Speci cation. Available on the internet: www.gs1.org.
Heng, Stefan: Heng, Stefan,RFID Chips: Enabling the E cient Exchange of Information(February 6, 2009). Deutsche Bank Research Paper. Available at SSRN: http://ssrn.com/abstract=1339543. Deutsche Bank Research Paper, February 6 2009. Available at SSRN http://ssrn.com/abstract=1339543. [KBM+00] Kindberg, Tim, John Barton, Jeff Morgan, Gene Becker, Debbie Caswell, Philippe Debaty, Gita Gopal, Marcos Frid, Venky Krishnan, Howard Morris, John Schettino and Bill Serra: People, Places, Things: Web Presence for the Real World. WMCSA 2000, Monterey, USA, December 2000.
Leaver, Sharyn: Evaluating RFID Middleware. port, Forrester Research, Inc., 2004. Available on http://www.bauer.uh.edu/r d/ForresterRFIDwave.pdf (18.02.2008).
Technical Rethe Internet: [Sch02] [Sun05] [ZGP03]
Leong, Kin Seong, Mun Leng Ng and Daniel W. Engels: EPC Network Architecture. Technical Report, Auto-ID Labs, Massachusetts Institute of Technology and Adelaide, 2005.