MALMÖ UNIVERSITY 205 06 MALMÖ, SWEDEN WWW.MAH.SE isbn 978-91-7104-607-9 (print) isbn 978-91-7104-608-6 (pdf)
ÅSE JEVINGER
TOWARD INTELLIGENT GOODS:
CHARACTERISTICS, ARCHITECTURES
AND APPLICATIONS
S TUDIES IN C OMPUTER SCIEN CE N O 1 , DOCT OR AL DISSERT A TION ÅSE JEVIN GER MALMÖ UNIVERSIT Y 20 1 4 T O W ARD INTELLIGENT GOODS: C HAR A CTERIS TICS, AR C HITECTURES AND APPLIC A TIONST O W A R D I N T E L L I G E N T G O O D S : C H A R A C T E R I S T I C S , A R C H I T E C T U R E S A N D A P P L I C A T I O N S
Studies in Computer Science no 1
© Åse Jevinger, 2014
ISBN 978-91-7104-607-9 (print) ISBN 978-91-7104-608-6 (pdf) Holmbergs, Malmö 2014
ÅSE JEVINGER
TOWARD INTELLIGENT GOODS:
CHARACTERISTICS, ARCHITECTURES
AND APPLICATIONS
Malmö högskola, 2014
Teknik och samhälle
The publication is also electronically available at: http://dspace.mah.se/handle/2043/17810
i
ABSTRACT
In the context of globalization, the requirements on transport logistics continuously rise. Often goods travel through many different countries, using several transport modes and involving a number of different actors. Implementing some level of intelligence on the goods, which provide them with the capabilities to assist in the logistical activities, is one of the instruments that can be used to improve control and efficiency in transports and goods-handling. The concept of intelligent goods both opens up for new types of services and may be used to improve currently available services.
The research is mainly focused on the characteristics, possible architectures, and applications of intelligent goods systems. In this context, an intelligent goods system refers to a number of interacting components, e.g. on-board units, servers, and RFID tags, which together provide intelligent goods services. Intelligent goods refer to goods with a higher degree of intelligence than just providing the ID of the goods, and generally the concept involve information processing and/or storage on or close to the goods, acting on behalf of the goods throughout the whole transport. The purpose of the studies is to investigate how intelligent goods can be used to improve goods transports in terms of more efficient goods-handling as well as better control of the goods and the transportation process, but also in terms of more efficient information sharing, e.g. between different actors. This may in turn provide reduced costs, environmental impact and usage of infrastructure. The research is concentrated on the communication and processing of information before, during and after transport. Most of the research results are applicable to
ii
goods transport by any mode, whereas some of the research has an emphasis on road transport.
A framework is presented which can be used to describe intelligent goods systems, including the capabilities of the goods, necessary information entities related to the goods, as well as a number of primitive services that can be used as building blocks when creating more advanced intelligent goods services. Furthermore, a new approach to service description is proposed, which can be used to, amongst others, define an intelligent goods service and to perform architecture analyses. By identifying architectures corresponding to different service solutions, intelligent goods can be compared with other types of solutions, for instance more centralized approaches. In particular, different situations and services put different requirements on a system and the benefits of using intelligent goods vary. In order to investigate how intelligent goods may be applied in practice, two services have been examined in more detail: a dynamic shelf-life prediction service, and a consignment-level emission allocation service. These studies involve field tests, interviews and simulations. Finally, an investigation of how intelligent goods systems can be modelled as multi-agent systems is also included.
iii
ACKNOWLEDGEMENTS
I would like to express my deepest gratitude and sincere respect to my supervisors Prof. Paul Davidsson and Assoc. Prof. Jan Persson. They have both provided excellent guidance and encouraged me throughout the research process. I would also like to direct my gratitude to assistant supervisor Dr. Henrik Sternberg for his valuable support and advice as well as indispensable help with external contacts. Furthermore, I would like to acknowledge and thank the members of the Swedish Intelligent Transport Systems Postgraduate school (NFITS), who have encouraged me to contemplate and structure my research. Thanks also to my colleagues both at Blekinge Institute of Technology and Malmö University for inspiration as well as warm friendships. A special thanks to all my other friends as well, who have been there for me in good times and bad. Finally, I thank my wonderful family for their love and endless support - thank you for putting up with me and for being part of my life.
The research presented in this thesis has been funded by the Swedish Road Administration, Vinnova, the J. Gust. Richert Memorial Fund, Blekinge Institute of Technology and Malmö University.
Malmö, October 2014 Åse Jevinger
iv
PREFACE
This thesis includes the following articles:
I. Jevinger Å., Persson J.A. & Davidsson P. (2010). Analysis of transport services based on intelligent goods. Proceedings of the 22nd Conference of the Nordic Logistics Research Network (NOFOMA).
II. Jevinger Å., Davidsson P. & Persson J.A. (2011). A framework for agent-based modeling of intelligent goods. Agents in Principle, Agents in Practice, Lecture Notes in Computer Science, Vol. 7047, pp. 97-112, Springer. (Proceedings of the 14th International Conference on Principles and Practice of Multi-Agent Systems.)
III. Jevinger Å., Davidsson P. & Persson J.A. (2014). A metamodel for intelligent transport system services. (Revised version of paper XI, submitted for journal publication.)
IV. Jevinger Å. and Persson J.A. (2014). Consignment-level allocations of carbon emissions in road freight transports. (Submitted for journal publication.)
V. Jevinger Å. and Davidsson P. (2014). Toward dynamic expiration dates: an architectural study. (Extended version of paper XII)
VI. Jevinger Å, Göransson M. and Båth K. (2014). A field test study on a dynamic shelf life service for perishables. Proceedings of the 26th Conference of the Nordic Logistics Research Network (NOFOMA).
v
In papers I-V, I have been the main contributor, from problem specification, research planning and solution development to the analysis of results and writing. In paper VI, I have contributed with the idea, as well as the major part of the literature review (not supply chain information sharing), minor parts of the methodology, and the major part of the results presentation and analysis. Göransson and I contributed equally to the discussion and conclusion sections. Båth has contributed with mathematical models for shelf-life calculations based on bacterial growth.
The following papers are related to but not included in the thesis:
VII. Jevinger Å., Persson J. A., Davidsson P. & Lumsden K. (2009). Analysis of intelligent goods and local decision making. Proceed-ings of the 16th Intelligent Transport Systems World Congress. VIII. Jevinger Å., Davidsson P. & Persson J.A. (2010). Agent
based intelligent goods. Proceedings of the 6th International Workshop on Agents in Traffic and Transportation.
IX. Jevinger Å., Davidsson P. & Persson J.A. (2011). A method for identifying and evaluating architectures of intelligent goods services. Intelligent Distributed Computing V, Studies in Computational Intelligence, Vol. 382, pp. 237-242, Springer. (Proceedings of the 5th International Symposium on Intelligent Distributed Computing.)
X. Jevinger Å., Davidsson P. & Persson J.A. (2011). A method for identifying architectural solutions for potential intelligent goods services. Proceedings of the 23th Conference of the Nordic Logistics Research Network (NOFOMA).
XI. Jevinger Å., Davidsson P., Persson J.A., Mbiydzenyuy G. & Bakhtyar S. (2012). A Service Description Framework for Intelligent Transport Systems: applied to intelligent goods. Proceedings of the 5th European Conference on ICT for Transport Logistics.
XII. Jevinger Å. & Davidsson P. (2014). Toward Dynamic Expira-tion Dates: An Architectural Study. Dynamics in Logistics, Lec-ture Notes in Logistics, Springer. (Proceedings of the 5th Inter-national Conference on Dynamics in Logistics.) In press.
vi Paper I Paper II Paper III Paper IV Paper V Paper VI Paper VIII Replaced by
Paper VII Paper IX
Paper X Extended by Paper XI Replaced by Paper XII Extended by Influenced Influenced
Figure 1. Illustration of the relationships between the different papers.
vii
CONTENTS
ABSTRACT ... i ACKNOWLEDGEMENTS ...iii PREFACE ...iv INTRODUCTION ... 1 1 BACKGROUND ... 5 1.1 Goods transportation ...5 1.2 Enabling Technologies ...6 1.3 Software agents ...7 1.4 Intelligent goods ...8 1.5 Research gap...11 2 RESEARCH FOCUS ... 13 2.1 Research questions ...13 2.2 Limitations ...152.3 Related research fields ...15
3 METHODOLOGY ... 17
4 CONTRIBUTIONS ... 23
5 CONCLUSIONS AND FUTURE WORK ... 31
viii
PAPER I
ANALYSIS OF TRANSPORT SERVICES BASED ON
INTELLIGENT GOODS ...41 1 Introduction ...43 2 Basic services ...44 3 Information entities ...46 4 Update services ...48 5 Functions ...49 6 Relationships ...50
6.1 Relationships between services and information entities ...50
6.2 Relationships between services and functions ...52
7 Location of processing and information entities ...52
8 Conclusions ...55
References ...56
PAPER II A FRAMEWORK FOR AGENT-BASED MODELING OF INTELLIGENT GOODS ...59 1 Introduction ...59 2 Intelligent Goods ...60 2.1 Definition ...60 2.2 Capability Dependencies...62 3 Services ...63 4 Agents ...64 4.1 Information Entities ...64 4.2 Single Agents ...66 4.3 Multiple Agents ...68 4.4 Locations of Agents ...72 5 Conclusions ...73 References ...73 PAPER III A METAMODEL FOR INTELLIGENT TRANSPORT SYSTEMS SERVICES ...77
1 Introduction ...77
2 Related work ...78
2.1 ITS Service description ...78
2.2 Web service description ...79
3 A Metamodel for ITS Services ...80
3.1 Abstract Service Metamodel ...80
ix
3.3 Discussion ...85
3.4 Service Composition/Decomposition Analysis ...85
4 Application of the ITS Service Metamodel on a Potential Intelligent Products Service ...88
5 Discussion and conclusions ...92
References ...93
PAPER IV CONSIGNMENT-LEVEL ALLOCATIONS OF CARBON EMISSIONS IN ROAD FREIGHT TRANSPORTS ...97
1 Introduction ...97
2 Literature review ...99
3 Selected reference allocation principles ... 100
4 Information used for allocation ... 102
4.1 Levels of information availability ... 102
4.2 Information availability during transport versus after ... 103
5 New allocation principles ... 103
5.1 New allocation principles with information availability level 4 ... 103
5.2 New allocation principles with information availability levels 2 and 3 ... 104
6 Simulation setup ... 105
7 Simulations without compensation factors ... 106
7.1 Results and analysis of simulation of one transport route ... 106
7.2 Results and analysis of simulations of 100 000 transport routes ... 109
8 Simulations with compensation factors ... 110
9 Summary and conclusions ... 112
References ... 113
PAPER V TOWARD DYNAMIC EXPIRATION DATES: AN ARCHITECTURAL STUDY ... 119
1 Introduction ... 119
2 Related Work ... 120
2.1 Concepts Related to the Expiry Date Service ... 120
2.2 Relevant Technologies ... 121
3 Methodology ... 122
4 Service Description ... 123
x
6 Target User Group Requirements ... 125
7 Heuristics ... 126
7.1 Basic Architectures ... 126
7.2 Capabilities and Sensors ... 127
8 Analytic Hierarchy Process ... 129
9 Conclusions and Future Work ... 131
References ... 132
Appendix: Questionnaire ... 134
PAPER VI A FIELD TEST STUDY ON A DYNAMIC SHELF LIFE SERVICE FOR PERISHABLES ... 139
1 Introduction ... 140
2 Related work ... 141
2.1 Supply chain information sharing ... 141
2.2 Cold chain monitoring ... 141
2.3 Remaining shelf life prediction ... 142
3 Methodology ... 143
3.1 Interviews ... 144
3.2 Field tests ... 144
3.3 Data analysis ... 145
4 Results and analysis ... 145
4.1 Interviews ... 145
4.2 Field tests ... 147
4.3 Discussion... 150
5 Conclusions ... 151
1
INTRODUCTION
The transport sector represents a significant economical factor, with new challenges and opportunities constantly arising in terms of globalization, regulations, production strategies, and advancements in information and communication technologies. The amount of freight transports is continuously growing, enabling economic growth, job creation and local regions to be integrated in the world economy (European Commission 2011). The increased traffic load, however, causes congestion, environmental damage, increased infrastructure wear etc. These problems are particularly evident in the road transport system. An increased use of transport modes other than road, for instance by using intermodal transport solutions, would improve sustainability (Natvig et al. 2009; Bektas & Crainic 2008). In intermodal transport solutions, two or more modes of transportation are used for delivering goods from origin to destination (Goel 2009). However, intermodal transports often require more complex logistic solutions involving a number of different actors which need to coordinate and exchange information. Similarly, the logistics requirements on unimodal transportations continuously increase as well. The goods often travel through several countries and different actors may be involved, e.g. different subcontracted hauliers (European Commission 2011; Unnikrishnan & Figliozzi 2011; Juga & Juntunen 2011). Efficient information sharing related to freight control and handling is thereby needed between the actors. The currently used information systems are typically based on centralized databases, and representing product-level information and communicating it between organizations can be a challenge, especially in case of mass-customization of products (Meyer et al. 2009).
2
Another challenge within the logistics domain is that despite several research reports on positive effects on collaboration between different market players, there is a general scarcity of successfully implemented collaboration schemes (Fugate et al. 2009). Lack of collaboration may lead to fragmented transport planning and control, which in turn causes inefficient transport executions, for instance in terms of low vehicle filling rates (Rodrigues et al. 2008; McKinnon & Edwards 2010). Given the current level of freight transport demands and low vehicle filling-rates in combination with transport costs, the environmental impacts and the limited road/rail capacities, transport efficiency needs to be improved (Baalsrud Hauge 2011; McKinnon & Edwards 2010; McFarlane et al. 2013). One way of improving transport efficiency and collaborations is to utilize Information Communication Technology (ICT) (Sternberg et al. 2013; Baalsrud Hauge 2011).
Since a lot of activities, and corresponding decisions, are performed locally in a transport chain (e.g. loading, unloading and priority handling) locally implemented services may represent a natural and motivated approach in both unimodal and intermodal transportations. In particular, such services may provide local access to information. Furthermore, they can often be made less dependent of a communication link between back-office and the goods or vehicle, and thereby less sensitive to problems caused by communication delays or lack of wireless coverage during parts of the transport (Jedermann & Lang 2008). The associated communication costs can also be reduced. For instance, a locally implemented monitoring service with data preprocessing capabilities might restrict the data transmission to warning messages, which are sent only when the quality falls below an acceptance limit or if a pending quality loss can be predicted (as opposed to a solution sending monitoring data continuously) (Jedermann & Lang 2008). Other advantages with locally implemented services are reduced decision lags, increased local dynamic solving of disturbances and an ease of complexity with an increasing number of products to be managed individually (Trentesaux et al. 2013).
The concept intelligent goods generally implies that some level of information processing and/or storage is implemented on, or close to the goods, acting on behalf of the goods throughout the whole
3
transport (Kärkkäinen et al. 2003; McFarlane et al. 2003; Wong et al. 2002). This intelligence may be used to, for instance, detect deviations (route, delay, temperature etc.), collect information about the context and conditions around the goods, and to inform about the goods specific requirements (destination, temperature, expected arrival time etc.) (Meyer et al. 2009; McFarlane et al. 2013). Delays may, for example, be detected in one part of the transport chain and enable appropriate actions in the remaining part. Intelligent goods are hereby capable of providing automated coordination and information sharing between the different actors. Furthermore, the ability to inform about the goods requirements and the ability to trace information related to the transport, for instance who was responsible when the goods were damaged or delayed, might lead to an increased trust between different actors (Forcolin et al. 2011). The increased trust may in turn enable increased collaboration with higher vehicle filling rates as a result (McKinnon & Edwards 2010). Additionally, apart from enhancing current transport services and operations, intelligent goods open up for new types of services as well. Two such services are presented in this thesis: a product level carbon auditing service leveraging the transport emissions caused by the product, and a dynamic shelf life service for perishables, based on local temperature monitoring.
Intelligent goods should be seen as a complement to a centralized approach which may make the information and information exchange better in some way. Furthermore, intelligent goods allow for services that would be difficult and/or costly to implement in back-office, for instance local monitoring of the conditions of the goods, including when they are placed outside typically monitored rooms, e.g. vehicles or warehouses. A higher level of flexibility to react to unforeseen events may also be enabled. However, the distributed approach has potential drawbacks in terms of sub-optimization and high costs (Jedermann & Lang 2008; Meyer et al. 2009; McFarlane et al. 2013). For instance, some studies show positive results from centralized transport planning (Janssen 2004), which indicates that this might not be a service suitable for intelligent goods (though this assumption may certainly change with the nature of the transport system). The level of costs associated with intelligent goods naturally depends on the required level of intelligence, where it is placed, and the amount
4
of goods items. It is also necessary to keep in mind that intelligent goods may reduce costs as well, thanks to the benefits provided by the concept (Jedermann & Lang 2008).
The research presented in this thesis focuses on intelligent goods as an instrument to improve goods transports in terms of more efficient goods-handling as well as better control of the goods and the transportation process, but also in terms of more efficient information sharing, e.g. between different actors. This may in turn provide reduced costs, environmental impact and usage of infrastructure, for instance if higher filling rates can be achieved. Most of the research results are applicable to goods transport by any mode, whereas some of the research has an emphasis on road transport. The primary motive for our research is that we believe that intelligent goods, despite some negative aspects, can be of value in transport logistics – a view that seems to be shared by the majority of the research papers written on the subject.
5
1 BACKGROUND
This section provides an overview of the problem domain and the background to this research. First, a description of goods transportation in the context of intelligent goods is given. Thereafter, a few of the most dominant enabling technologies for intelligent goods are described. Furthermore, since a large number of the studies related to intelligent goods suggest using software agents as enablers for a higher level of intelligence of the goods (see section 1.4), a short review of agents has also been included. Then, a survey of the intelligent goods concept and the current state of the art is presented. The section ends with the identified research gap.
1.1 Goods transportation
The number of parties involved in a logistics solution varies depending on the type of solution applied. Ho and Lim (2001) describe, in a Morgan Stanley report, the characteristics of the different types of logistics service providers. A first-party logistics (1PL) provider, often a small manufacturer buying and selling in the same location, essentially owns and handles all logistics functions himself. A second-party logistics (2PL) provider offers a small number of supply chain functions, such as trucking or warehousing. A third-party logistics (3PL) provider is involved in the management of the supply chain and performs a large proportion of a client’s supply chain logistics activities. Freight forwarding or contract logistics companies are 3PL providers. A fourth-party logistics (4PL) provider is essentially a logistics integrator, responsible for contracting various 2PL or 3PL providers. Finally, fifth-party logistics (5PL) providers focus on providing e-logistics solutions for the entire supply chain. In general,
6
the more actors involved in the transport chain, the more complex the information exchange between the different actors often becomes. Furthermore, the need for monitoring and tracing information related to the transport may also increase with the number of actors. Since these types of activities are performed locally, decentralizing such services and information by using intelligent goods, might be a motivated solution.
Goods may be transported by means of different traffic types, such as line traffic, dedicated transports and chartered transports (Lumsden 2006). Line traffic follows fixed timetables and routes, and uses intermediate terminals for consolidation, collection and distribution. Dedicated transports can be seen as a kind of private line traffic based on the requirements from only one customer which produces large and stable flows of goods. Finally, chartered transports are single, usually door-to-door, transports induced by customer orders. One of the more visionary identified services related to intelligent goods is to let the goods find its own way through the transport network (Scholz-Reiter et al. 2004). The line traffic type would typically suit such a service.
Intelligent goods may be implemented using the primary package, or the secondary, tertiary etc., or the load carrier, as a holder of local intelligence acting on behalf of the goods. Usually standardized load carriers are employed within transportation in order to facilitate the handling and optimize the loading. New, reusable load carriers, for instance pallets, have also entered the market and these are particularly suitable for intelligent goods since they can be reused and thus reduce the costs.
1.2 Enabling Technologies
Intelligent goods may involve the use of, for instance, Radio Frequency Identification (RFID) technology, Wireless Sensor Networks (WSN) and/or printed electronics. RFID is commonly used for automatically identifying and tracking objects. In its basic form (passive tags), RFID has the potential to replace the currently used bar codes since it provides more efficient reading, tracking and real-time transparency (Danish Technological Institute 2014). The more advanced RFID tags (active tags) incorporate a battery and they may moreover initiate communication. The RFID technology has weaknesses though which
7
may prevent 100% readability. This thesis has a slight emphasis on RFID, however, since new technologies may arise in the future, no assumptions are made in most of the included work, on which underlying technology is used for implementing intelligent goods.
WSN consist of sets of autonomous sensors, which monitor physical conditions, for instance temperature. In contrast to RFID tags which communicate directly to a reader, WSNs allow multihop communication. A number of standards and technologies are available for WSNs, such as Bluetooth, ZigBee and WiFi (Rawat et al. 2013). RFID and WSN may be integrated to provide solutions that use RFID for identification and data processing, and WSN for sensing.
An emerging alternative to the traditional silicon-based electronic components is organic electronics, which are based on electrically conductive polymers (Chaves & Decker 2010). These circuits and considered easier to develop since they may be printed using standard industrial printers, and they are furthermore lighter and less expensive than traditional electronic components. The performance with respect to conductivity and reliability is, however, relatively worse (Chaves & Decker 2010). Research efforts continuously push the advancement of organic electronics, and components such as displays, sensors and RFID tags are being developed (Marien et al. 2013; Tehrani et al. 2010).
1.3 Software agents
’Software agent’ can be seen as an umbrella term covering a number of different agent types, such as intelligent agents, autonomous agents, distributed agents, mobile agents, multi-agent systems etc. These concepts focus on different properties of the agents and the general definition of software agents has thereby been subject to discussions. None of the definitions suggested over the years has reached world-wide consensus but one of the most popular ones reads as follows (Wooldridge & Jennings 1995):
“An agent is a computer system that is situated in some environment, and that is capable of autonomous action in this environment in order to meet its designed objectives.”
Agents can usefully be applied to the area of intelligent goods for a number of reasons. Primarily, both agents and intelligent
8
goods are grounded on the vision of autonomous entities that acts on behalf of something or someone. Furthermore, one of the fundamental capabilities of agents is that they should detect and react to changes in the environment. Agents hereby represent a natural, though not exclusive, way of modelling intelligent goods on a conceptual level. The higher the levels of intelligence of the goods are, the more apparent are the benefits of modelling intelligent goods as agents. Intelligent goods agents act on behalf of the goods and they furthermore enable the goods to interconnect and coordinate, for instance to make sure that only goods allowed to be grouped together are loaded onto the same vehicle. In general, agents as well as intelligent goods allow for autonomous solutions which decrease the need for human intervention.
1.4 Intelligent goods
Various denotations for concepts similar or identical to intelligent goods can be found in literature, including Intelligent cargo (Euridice 2011; Huschebeck et al. 2009), Smart goods (Holmqvist & Stefansson 2006), Smart freight (Lumsden & Stefansson 2007) Intelligent products (Wong et al. 2002), Intelligent packaging (Johansson 2009), etc. Some of the concepts only differ slightly in their definitions; some of them are not properly defined in literature whereas others have been defined differently in different studies. In particular, the level of intelligence differs in different studies. Due to the ambiguity around the actual meaning of the concept intelligent goods, a separate research question of this thesis has been dedicated to the area.
High-value goods are often allowed higher transportation costs than low-value goods (Lumsden 2006). Similarly, time critical deliveries (or deliveries critical in some other way) might also motivate high transportation costs, irrespective of the actual value of the goods. In general, the most potential application area for expensive and sophisticated solutions, including for instance a high level of intelligence implemented on the goods, are thus within the transportation of valuable or (time) critical goods. However, it is important to note that the concept intelligent goods does not necessarily imply implementing intelligence on the actual unit of goods. The intelligence might just as well be implemented somewhere close to the goods, as long as it is present throughout the whole
9
transport and acts on behalf of the goods, i.e. acts as an enabler through which the goods may, for instance, make decisions. The actual goods items might instead, for instance, only be supplied with less intelligent tags storing information about ID, decision rules etc.
The simplest and often least expensive solutions related to intelligent goods involve locally stored information about the goods (ID, delivery instructions, necessary handling instructions, etc.) and some means for this information to be communicated to the information systems in the supply chain (Kärkkäinen et al. 2003). Usually these solutions are based on RFID tags. Goods equipped with more advanced capabilities have so far been more sparsely investigated, though the number of studies continuously rises. Meyer et al. (2009) classify Intelligent Products from different aspects. In particular, they identify two extreme approaches concerning the location of the intelligence: intelligence through network and intelligence at object. Intelligence through network implies that the intelligence is placed outside the physical product, for instance as a dedicated agent on a server. The agent is then combined with a device on the product, which is only used as an interface to the intelligence (Främling et al. 2003; Meyer et al. 2014). Intelligence at object indicates that all intelligence takes place at the physical product (Carabelea et al. 2003). Furthermore, the level of aggregation of the intelligence is also categorized by Meyer et al. (2009). They state that an object may either be an intelligent item, which means that it only manages information, notifications and/or decisions about itself, or it may be an intelligent container, which manages itself as well as the component it is made of and may act as a proxy device for them.
Most of the research papers related to intelligent goods can be classified as conceptual papers whereas empirical approaches are less common (Sternberg & Andersson 2014). Many of the studies focus on the logistics within one facility (e.g. a manufacturing system or a warehouse) (Van Belle et al. 2011; Scholz-Reiter et al. 2010). These are out of the scope of this thesis; however, they have been used as a source of inspiration. Some of the studies focus on frameworks (Langer et al. 2006; Stefansson & Lumsden 2009). For instance, Stefansson and Lumsden (2009) present a framework that is based on three cornerstones: smart freight, smart vehicle and smart infrastructure. Amongst others, the framework aims at
10
improved information accessibility by using RFID-tagged goods carrying extensive data. Thereby local decision-making is enabled as a complement to centralized planning and decision-making. A number of proposals exist concerning the interface for collecting and exchanging information between intelligent products and with other information systems (Främling et al. 2013).
A large number of the studies related to intelligent goods suggest using agents as enablers for a higher level of intelligence of the goods (Schumacher et al. 2009; Jedermann & Lang 2008; Langer et al. 2006; Främling et al. 2013). Goods equipped with agents might entail relatively high costs. However, Jedermann & Lang (2008) state that the increase in hardware costs for additional computation power for local data analyses and decision-making, is moderate in comparison to the primary costs of active wireless sensors and telematics units. They furthermore claim that the application field of intelligent packages is mainly limited to settings where a high amount of local information has to be processed, e.g. involving sensors.
A number of trials and projects involving different levels of intelligence of the goods have during the recent years been conducted within the transport sector (Huschebeck et al. 2009). The EU project Intrans (Intrans 2009) works with the vision to: “Enable a fully automated, multimodal and environmentally friendly freight transport system, where the goods find the most efficient way through the supply chain, based on information and communication technology, and advanced control models and decision support”. Intrans proposes a separate role called Intelligent Goods as an extension to the Arktrans reference model (Natvig et al. 2009). The main responsibilities of this role would be to store and protect information linked to the goods, monitor the transport of the goods in relation to limitation attributes (e.g. planned and stored route), store and/or send messages, and to communicate with the transport environment of the goods, e.g. roadside equipment or gate equipment in terminals. Intrans furthermore identifies the Transport Execution Plan (TEP), introduced by the Freightwise project (Freightwise 2008), as a very relevant data set in relation to intelligent goods. The TEP is a plan established between a Transport User and a Transport Service Provider (Fjørtoft et al. 2009), which include information about
11
sender, receiver, earliest and latest delivery times, details about the goods, transport instructions etc.
Euridice (Euridice 2011) is a project funded by EU, which focuses on the individual cargo item and its interaction with the surrounding environment as well as the user. According to their vision, intelligent cargo should be able to connect itself to logistics service providers, industrial users and authorities to exchange transport related information and perform specific services whenever required along the transport chain (Schumacher et al. 2010). Euridice believes that in the upcoming years, the usage of passive and active RFID will increase due to price reductions, which will lead to a situation where more and more local intelligence is available. The abilities for local information collection and processing as well as local decision-making on the basis of this information will hereby also be improved. Furthermore, Euridice predicts these abilities not only to be present on vehicle and container level, but also on pallet and goods level in order to allow for a greater flexibility to control and steer the transportations of goods (Schumacher et al. 2010). This view is however not shared by everyone (Huschebeck et al. 2009; Sternberg & Andersson 2014). Large parts of the research community focus more on cheaper, passive tags since they regard these as more promising. However, the prices for more advanced tags are continuously decreasing and the effects of this remain to be seen. Furthermore, new technologies, such as printed electronics, evolve, which are considered both lighter and less expensive than traditional electronic components. According to Das and Harrop (2013), the RFID market will increase from nearly $8 billion in 2013 to over $30 billion in 2024. Both active and passive RFID markets are expected to grow, and with the increased markets, the prices will most probably drop (Das & Harrop 2013).
1.5 Research gap
As mentioned above, many of the existing research papers related to intelligent goods, and similar concepts, can be classified as conceptual papers. Nevertheless, some vital parts, preventing ad-hoc solutions, are still missing in this conceptual research. For instance, in order to be able to implement intelligent goods services in a structured and well-considered manner, there must be a way to identify and appraise different solution alternatives. Furthermore, to find out what
12
intelligent goods have to offer and which services can be combined, the possible intelligent goods services should be compiled and their requirements be investigated. The research community would also benefit from a common view on the definition of intelligent goods. Therefore, even though many of the existing research papers are on a conceptual level, which means that empirical approaches are less common (Sternberg & Andersson 2014), a lot of the research presented in this thesis is on a conceptual level as well, aiming to reduce the missing parts. The objective has been to provide a solid conceptual basis, which facilitates further empirical studies.
We have, however, also carried out some empirical studies, aiming at pushing this less explored research area forward.
13
2 RESEARCH FOCUS
This section presents the research focus by describing the research questions, the limitations of the research, and by describing how the research relates to other research fields.
2.1 Research questions
The purpose of the studies is to investigate how intelligent goods can be used to improve goods transports in terms of more efficient goods-handling as well as better control of the goods and the transportation process. Moreover, the concept is also studied with the aim to improve information sharing, both between different actors (or to the final consumer), and between the goods and back-office. The research is concentrated on the communication and processing of information before, during and after transport. Goods transport by road, rail, sea and air are in focus, with a slight emphasis on road. Local movements of goods within a facility are not included.
The thesis addresses the following research questions:
RQ 1 How can intelligent goods systems be described and char-acterized in a way that enables further analysis?
To address the subsequent research questions, it is neces-sary to have a clear picture of the main characteristics of intelligent goods. For instance, there are several levels of intelligence; from goods only knowing its own identity to more advanced levels of intelligence, such as autonomous decision-making.
14
RQ 2 Which are the services that intelligent goods systems may provide and what are the requirements of these services?
The primary aim with this research question is to identify possible intelligent goods services. Furthermore, by investi-gating these services, more knowledge about which capa-bilities are required by intelligent goods can be gained, for instance in terms of common information entities and functions.
RQ 3 How can software agents be u2.2sed in the development of intelligent goods systems?
Since software agents are often suggested as enablers for a higher level of intelligence of the goods (see section 1.4), it is important to provide an understanding of how intelligent goods service can be modeled as software agents, based on established agent modeling techniques.
RQ 4 How can intelligent goods system architectures for a certain service (or set of services) be identified?
This research question is focused on which information processing is needed, where the information should be stored and which communication links are required. The aim is to identify architectures ranging from placing all intelligence on the goods level to placing all intelligence on a central level, in order to be able to evaluate the different alternatives.
RQ 5 How can different intelligent goods system architectures be compared, and when are centralized or distributed solutions most beneficial?
In order to determine the advantages and disadvantages with intelligent goods, it is necessary to compare different alterna-tives to each other. Both different solutions based on intelli-gent goods, and distributed versus centralized configurations need to be compared. This research question focuses on how such analyses can be performed, and when distributed solu-tions are beneficial compared to more centralized solusolu-tions.
15
RQ 6 How can the concept of intelligent goods be applied in practice, and what are the potential and limitations of such services?
This research question focuses on how, and for which services, intelligent goods may be applied in practice, involving e.g. field studies and simulations of intelligent goods systems. The potential and limitations, primarily in terms of feasibility, need and information quality, of using intelligent goods in such services are also investigated.
2.2 Limitations
This thesis does not consider the business aspects of implementing intelligent goods, in terms of actual costs and revenues, since this would take a lot of efforts from the identified research questions (and could in fact be a PhD thesis in itself). In particular, the total costs of an implemented intelligent goods solution can be calculated by summing up the costs of devices and manpower. However, the profits are usually harder to estimate (Wang & Ip 2013). Furthermore, we have aimed at keeping the research independent of brand-specific solutions. Therefore we have not considered the practical limitations associated with different tags and readers available on the market today, which may restrict the utility. Finally, the scope of the thesis is limited to the use of intelligent goods in transportation.
2.3 Related research fields
The research presented in this thesis touch on multiple disciplines, such as logistics, computer science and information systems (see methodology section). Both within logistics and computer science, the term Smart/Intelligent Products has been used to denote objects with enhanced capabilities, for instance in terms of sensing, process-ing and networkprocess-ing capabilities (Gutierrez et al. 2013). In logistics, intelligent goods and similar concepts are part of a larger research area, incorporating smart/intelligent containers, infrastructure, ve-hicles etc., aiming for a smarter supply chain (Stefansson & Lums-den 2009; Jedermann & Lang 2008). Within computer science, and distributed intelligence in particular, the intelligent goods concept has many similar counterparts. Ubiquitous Communication implies a general ability of objects to communicate (anywhere and anytime),
16
Pervasive Computing focuses on objects enhanced with processing power (the environment around us becomes the computer), and Am-bient Intelligence addresses objects with the capability of registering changes in the physical environment and thus actively interacting in a process (Dohr et al. 2010). The main idea behind these concepts is that almost any object can be provided with enhanced capabili-ties, which allows the user to interact with the objects. However, the concepts tend to focus more on how human users interact with their environment, than the intelligent goods concept does (Meyer et al. 2009). In that sense, the Internet of Things (IoT) is more similar to intelligent goods. IoT is a concept based on the idea that a number of things around us, such as sensors, actuators, mobile phones, etc., will be able to interact and cooperate with each other, by having a locatable, addressable, and readable counterpart on the Internet (Atzori et al. 2010; Roman et al. 2011). This scenario requires the integration of several technologies and communication solutions. A number of application areas related to intelligent goods can be iden-tified within the scope of the concept, for instance goods equipped with tags and sensors sending information to traffic control sites, or pervasive computing in conjunction with sensors for the monitoring of goods condition status, e.g. temperature, humidity or shock (At-zori et al. 2010; Schumacher et al. 2011). The IoT concept, howev-er, seems to be more concentrated on connectivity and information exchange than on the actual intelligence of the products (Meyer et al. 2009). Finally, yet another concept closely related to intelligent goods is Artificial Intelligence, and software agents, as described in section 1.3.
17
3 METHODOLOGY
The research presented in this thesis primarily belongs to information systems design science. Design science within information systems “seeks to extend the boundaries of human and organizational capabilities by creating new and innovative artifacts” (Hevner et al. 2004). It was selected as primary research approach since it is an applied research discipline directed towards creating applicable solutions to problems, by developing artifacts, at the intersection of IT and organizations (Peffers et al. 2007). The research presented in this thesis has resulted in a number of artifacts aimed at finding applicable solutions for how to improve the efficiency, control and information sharing in goods transportation, by means of intelligent goods. Apart from information systems design science, the research has also been influenced by game theory and systems theory.
Hevner et al. (2004) present a framework for understanding, executing, and evaluating information systems (IS) research. The framework includes the environment, which defines the problem space and the business needs, and the knowledge base, which provides the raw materials to the IS research (theories, frameworks, etc.). The business needs and the knowledge base form the input to the IS research process. The principal business needs for our research have been described in the introduction. The knowledge base is composed of conceptual modelling (Leshem & Trafford 2007), computational modelling, and existing expertise concerning intelligent goods, field tests and interviews. The objectives with the interviews were to get information from domain experts that could be used both as input to the research, and to confirm the relevance of the research.
18
During the research process artifacts, defined as constructs, models, methods and instantiations, are built and evaluated (Hevner et al. 2004). Building and evaluating artifacts represent the research activities whereas the resulting constructs, models, methods and instantiations describe the research outputs (March & Smith 1995). According to Peffers et al. (2007), a design research artifact can actually be any designed object in which a research contribution is embedded in the design. The artifacts specified by Hevner et al. (2004) are as follows. Constructs provide the language, in terms of vocabulary and symbols, in which problems and solutions are defined. Models use constructs to abstract and represent a real world situation. Methods define the processes by algorithms and practices, providing guidance on how to solve problems. Finally, instantiations show that the constructs, models, or methods can be implemented in a working system, for instance by creating a prototype system. The research in this thesis is primarily on a conceptual and theoretical level, and therefore the main focus has been on constructs, models and methods. Some research outside design science has also been conducted, such as investigations of the potential and limitations of intelligent goods services. These investigations are based on simulations and interviews.
The artifacts presented in the thesis have been evaluated in order to demonstrate their accuracy and usability. Depending on the nature of the artifact, evaluation may focus on functionality comparisons of the artifact in relation to the defined objectives, quantitative performance measures, satisfaction surveys, client feedback, simulation results etc. (Peffers et al. 2007). Hevner et al. (2004) list five concrete evaluation methods: observational, analytical, experimental, testing, descriptive. In this thesis, we have primarily applied observational, experimental and descriptive evaluation. Observational evaluation refers to case and field studies, experimental evaluation involves controlled experiments and simulations, and descriptive evaluation is performed using informed arguments and scenarios.
Corley and Gioia (2011) present results indicating that theoretical contributions currently rest largely on the ability to provide originality and utility. Within utility, top-tier journals seem to primarily reward those papers demonstrating theoretical contributions that are scientifically but not pragmatically useful. Hevner et al. (2004), on
19
the other hand, claim that the goal of design science research in IS is utility, and that scientific research should be evaluated in light of its practical implications. In this thesis, both contributions aiming at practical utility and scientific utility are presented.
The work described in this thesis can be seen as following a bottom-up approach, see Figure 2. We start at the lowest level with the question of how to describe and characterize intelligent goods systems, including information entities and functions required, as well as the capabilities of intelligent goods (RQ1). These can be used to describe possible intelligent goods services, and their requirements, as well as to explore agent technology in combination with intelligent goods (RQ2, RQ3). Based on some of these findings, alternative locations of the information and processing (which determine the required capabilities) as well as evaluation possibilities were examined (RQ 4, RQ5). Methods for how to identify a suitable architecture were also developed. Finally, the intelligent goods concept was applied to two concrete applications (RQ6). The main approach can thereby be seen as bottom-up. However, the lower blocks in Figure 2 have continuously been refined based on new findings in the upper blocks.
Capabilities of intelligent goods systems
Information entities and functions required in
intel-ligent goods systems
Formal specification of in-telligent goods services Characteristics of Intelligent goods systems
Possible intelligent goods services
Agent technology applied for intelligent goods
ser-vices Intelligent goods systems
architectures
Evaluation of intelligent goods systems
architec-tures Services and agent methodology
System architectures Consignment-level
alloca-tions of carbon emissions in road freight transport
Dynamic shelf life predic-tion Concrete applications
Figure 2 Research approach
20
All literature studies described in this thesis are based on traditional, narrative reviews performed through database searches and snowballing (backward and forward). Scientific journal articles and conference papers were primarily selected; however, other sources of information were also studied, such as academic theses, books and project reports.
As a first step towards an answer to RQ1, literature studies on intelligent goods and other related concepts were performed. Both definitions and the context, in which the goods are intended to operate, were studied. Since the concept of software agents bear similarities with the concept of intelligent goods (Meyer et al. 2009), the characteristics of the former were also studied and compared to the definitions and capabilities of intelligent goods. Additionally, visits to workshops, exhibitions and a RFID laboratory and test center (Danish Technological Institute 2014) were made. The visit to the test center provided insights regarding the capabilities and limitations of different types of silicon-based RFID tags.
The work concerning RQ1 also involved literature studies of different service definitions, related to various fields of research. The purpose of these studies was to gain knowledge and ideas of how a possible intelligent goods service might be specified. The results of this work formed the basis for a new ITS service metamodel, which can be used to uniquely specify ITS services.
RQ2 involves an investigation of which services might be implemented using intelligent goods. Some of the services were identified through studies of the transport chain and of both currently available and future ITS services related to goods transports, whereas others were presented as possible intelligent goods services in literature. The investigation resulted in a list of possible intelligent goods services, which not only served as input to RQ2 but were actually used in the processing of all the subsequent research questions. Furthermore, it provided more knowledge about which capabilities, information entities and functions are actually required by intelligent goods.
RQ3 was approached mainly from the perspective of how to model intelligent goods as software agents. The benefits of using agent modelling have proved to be more pronounced with more advanced intelligent goods services. However, in order to enable a clear and
21
lucid analysis, relatively primitive services were used as illustrative examples. Two different modeling methods were applied; Agent Unified Modeling Language (AUML) and Gaia (Odell et al. 2001; Wooldridge et al. 2000). AUML was chosen because it is a well-established modeling approach for multi-agent systems, supported by the Foundation for Intelligent Physical Agents (FIPA), and because it allows both inter- and intra-agent behavior descriptions. The inter-agent descriptions were used to describe the multi-agent communications whereas the intra-agent behavior descriptions were used to investigate the required functionalities and capabilities of the agents. Furthermore, Gaia provided the instruments for the system analysis and high-level design of the agents, and the agents resulting from the Gaia methodology were subsequently described in more detail using AUML. This process benefitted from the organizational design approach of Gaia.
All studies related to RQ1–3 were kept independent of the actual location of service information and processing. The question of whether to base the services on intelligent goods or not, was thereby kept open at this stage of research.
The main focus of the work with RQ4 was to investigate how possible intelligent goods services can be decomposed and how these different components can be distributed across different parts of the system. The aim behind this was to identify solutions ranging from placing all intelligence on the goods level to placing all intelligence on a central level, in order to be able to value the different alternatives against each other. Analysis and literature studies on service descriptions formed the basis for this research.
Analyses as well as literature studies on quality attributes and analytical approaches formed the basis of the work with RQ5. In particular, the Analytic Hierarchy Process (AHP) was used for comparing different architectures with each other. AHP provides an approach to select the most suitable alternative from a number of alternatives evaluated with respect to several criteria, and this approach was useful when evaluating the advantages and disadvantages with different architectures, under certain conditions. A questionnaire, which served as input to the AHP, was given to 5 supply chain actors within food production and logistics. The questionnaire provided views on the relative importance of different quality attributes,
22
with respect to the service, and was given in personal meetings or via email, complemented with explanations over the phone. In order to determine when centralized solutions are preferred, and when distributed solutions are preferred, an analysis of different service characteristics was also performed. Moreover, simulations were conducted to determine the effects of using intelligent goods in a dynamic transport planning system with limited information availability, as opposed to a system with non-limited information availability. The service being simulated allocates transport emissions shares to individual consignments (see RQ6 below).
As for RQ6, two concrete applications were used to illustrate how intelligent goods can be applied in practice: a dynamic shelf-life prediction (DSLP) service providing dynamic expiry dates for chilled food products, and the service allocating transport emissions shares to individual consignments. The work with the DSLP service involved semi-structured interviews which provided general views on a DSLP service. The interviews involved one customer organization and 10 supply chain actors within food packaging, food production, logistics, retail, and sensor technology. All interviews were conducted in personal meetings. Field tests were also performed in order to gain knowledge about the premises for the DSLP service. The work with the emission allocation service also included contacts, in this case in the form of discussions, with 3 logistics providers, which provided information about the level of aggregation of the allocation principles currently used within the organizations respectively. Simulations were also conducted in order to show how intelligent goods can be applied. The information from the transport actors, in combination with the results from the simulations, confirms the relevance of this type of service, when allocating emissions to consignment level.
23
4 CONTRIBUTIONS
The relationships between the papers included in this thesis and the research questions investigated are illustrated in Table 1. The subsequent text describes in more detail the contributions.
Table 1. Relationships between research questions and papers
Research question Papers
RQ1 I, II, III, V RQ2 I, II, IV, V RQ3 II RQ4 III, V RQ5 I, IV, V RQ6 IV, V, VI
RQ1: How can intelligent goods systems be described and characterized in a way that enables further analysis?
The literature studies on intelligent goods showed that the characteristics of intelligent goods (and related concepts) differed between the studies. In particular, only one or a few levels of intelligence are included in the definitions. Depending on the purpose of using intelligent goods, different levels of intelligence are needed and we believe that these levels should be reflected by the definition. Based on the literature reviews of intelligent goods, software agents and possible intelligent goods services, we propose a new intelligent goods definition, first presented in papers I and II, and then revised in paper V. The definition differs from others in that it captures several levels of intelligence reflecting the various types and levels of
24
capabilities the goods may possess. In practice these capabilities may be located on or close to the goods and they are present throughout the whole transport. The definition also suggests that all goods with capabilities above the lowest level in at least one of the dimensions should be called intelligent. Table 2 shows the proposed capability dimensions and the different levels.
Table 2. Capabilities of goods (paper V)
Dimension Capability
A. Memory storage 1. Ability to store ID
2. Ability to store goods data (other than ID) 3. Ability to store algorithms/decision rules
B. Memory dynamics 1. Static memory
2. Ability to change/add/delete data (other than ID) 3. Ability to change/add/delete algorithms/decision rules
C. Communication out 1. Data (including ID) can be read by external unit 2. Ability to send short-range communication messages 3. Ability to send long-range communication messages
D. Communication in 1. None
2. Data (other than ID) can be written by external unit 3. Ability to receive short-range communication messages
4. Ability to receive long-range communication messages
E. Processing 1. None
2. Ability to execute decision rules (e.g. If–Then statements)
3. Ability to execute algorithms (e.g. planning capability, optimization algorithms)
F. Autonomy 1. None
2. Reactive capability (actions must be triggered by an external unit)
3. Proactive capability (no external trigger needed)
G. Sensor 1. None
2. Sensor capability incl. ability to read sensor data (e.g. enclosed temperature sensors)
H. Time 1. None
2. Ability to measure time intervals 3. Ability to determine actual time
25
There are a number of advantages with letting the intelligent goods definition include different levels of intelligence. First of all it gives a more comprehensive description of the concept. Secondly, it may be used to classify intelligent goods applications, for instance based on different types of RFID tags. Furthermore, it may also be useful when mapping intelligent goods applications to the requirements of a set of intelligent goods services. For instance, when investigating which services can be supported by a specific application, the definition can be used to compare the capabilities of both parts in a structured way.
As a further complement to the studies related to RQ1, we identified a number of primitive functions and information entities (paper I), which can be used as a basis when creating intelligent goods services. These represent the low-level functionalities and information required by possible intelligent goods services, and the functions can be seen as a connecting layer between services and the information entities. The information entities are related either to the goods or to the housings1. A housing denotes the local entities surrounding or housing the goods, e.g. packaging layers, transport pallets, transport containers, vehicles, terminals, warehouses. The capabilities of the housings may serve several goods entities, however they may change during the transport, since loading and unloading may cause a change of the housing. Depending on the requirements of a particular service, the set of information entities in paper I can be complemented with additional information entities, as in papers II and III.
As a part of RQ1, the question of how to specify a possible intelligent goods service was also addressed. Based on literature studies and analysis, a new ITS service metamodel was developed, which can be used for describing and analyzing ITS services (Paper III). In particular, it can be used for architecture analyses (see RQ4 contributions below).
Finally, in order to relate the goods to its surroundings, a model of the context of the goods was developed in paper II, describing the main alternatives for placing information and processing required by the services.
1 In paper I the housings are named containers whereas the outmost container is called housing. This denotation was changed in paper II in order to avoid confusion with shipping containers, which only represent a part of what is intended.
26
In summary, the studies related to RQ1 resulted in a framework for describing intelligent goods systems which includes the following constructs (Hevner et al. 2004):
• Capability dimensions of goods (papers I, II and V) • Information entities related to goods and housings
(papers I, II and III)
• Primitive functions (Paper I) • ITS service metamodel (paper III)
It also includes the following model (Hevner et al. 2004): • Description of the context in which the goods function
(paper II)
The capability dimensions and information entities are evaluated in several of the papers, primarily in papers II, III, V (e.g. by scenarios) and the ITS service metamodel is evaluated in paper III. The evaluations are primarily descriptive (Hevner et al. 2004).
RQ2: Which are the services that intelligent goods systems may provide and what are the requirements of these services?
The main contribution related to this research question is a set of relatively primitive services, presented in paper I, which can be used as building blocks when creating more advanced intelligent goods services. The services inform about what goods to load or unload, estimated time of arrival, physical conditions or position of the goods etc. As a complement to the primitive services, an additional set of services was also identified. Since some of the information entities required by a service may need to be automatically and dynamically updated during the transport, a number of update services are also needed. In paper I, all update services needed by the listed primitive services, are presented. The two sets of services represent resulting constructs. Paper I also models the relationships between the services, including the update services, and the functions as well as between the services and the information entities. Apart from informing about the requirements of each of the services, these relationships naturally also show which services share the same requirements. These findings might be useful when investigating which services can be supported by a specific set of implemented functionalities, and vice versa.
27
The primitive services were used as building blocks for more advanced in several of the papers, in particular in papers II, IV and V. The more advanced services are also considered as constructs, e.g. the DSLP service and the mission allocation service. Descriptive evaluation and interviews were applied to confirm the relevance of some of the services.
RQ3: How can software agents be used in the development of intelligent goods systems?
The studied literature contained some investigations of how to use agents as an instrument for implementing solutions related to intelligent goods, for instance by combining advanced RFID technology with agents. However, the research knowledge appeared to be scarce. Paper II contributes to this area by investigating how intelligent goods services can be modeled as agents. The suggested modelling method (Hevner et al. 2004) is illustrated and evaluated by applying it to three specific services. The evaluation method is thereby descriptive. For each of the services, both single agent and multi-agent solutions are presented. The capabilities of all identified agents are furthermore mapped to the intelligent goods framework capabilities, in order to show which (distributed) capabilities are required by each of the agent solutions. By using Gaia and AUML, common functionalities were extracted from the services into separate agents, resulting in the identification of two “basic agents”. A comparison between the basic agents and the primitive functions, listed in paper I, shows that the one of the agents actually implements two of the primitive functions and the other agent implements three of the other primitive functions. An extraction of functionality, into basic agents, from a much larger set of services would possibly result in basic agents implementing the two remaining primitive functions as well. Like the primitive functions, all these basic agents can be used as a platform to build intelligent goods solutions on.
The level of decision-making and autonomy of some of the agents identified in paper II, as well as the amount of coordination between the agents, might seem a bit low to motivate the use of agents. The reasons for addressing these services are that we want to give simple and intelligible examples of how this type of analysis can be performed and how basic agents can be identified. The benefits of ‘agentification’ will naturally be more apparent with more advanced services.
28
RQ4: How can intelligent goods system architectures for a certain service (or set of services) be identified?
In order to be able to evaluate the advantages and disadvantages with intelligent goods in comparison to other alternatives, different architectures need to be identified and compared. Paper III presents a method, based on the ITS service metamodel, for identifying architectures for possible intelligent goods services. The metamodel is used to decompose a service. All possible locations of the resulting components can thereby be listed and the resulting communication and capability requirements can be identified. By decomposing a service, synergies with other services may also be discovered.
Paper V describes a method similar to above, which involves identifying all critical activities that must be performed, as well as the possible locations of each of these activities. Thereafter the most promising architectures can be selected, based on AHP. The method is applied to a service providing DSLPs for chilled food products. As a part of the results, a model of all possible communication paths within the service is provided.
The architecture analysis part of the ITS service metamodel is descriptively evaluated in paper III by applying it to an intelligent goods service. The approach taken in paper V is similarly descriptively evaluated by applying it to the DSLP service.
RQ5: How can different intelligent goods system architectures be compared, and when are centralized or distributed solutions most beneficial?
Depending on the circumstances and the functionality of a service, different architectures might be appropriate. Some services might require at least a minimum level of local intelligence whereas others are possible to implement completely on either a central or local level. For instance, the capabilities of the goods item versus the housings is naturally an interesting question both from a functionality, but also from a cost, perspective. In particular, it is necessary to compare different alternatives to each other, especially different intelligent goods solutions, in order to be able to determine the advantages and disadvantages with intelligent goods. In paper V, the identified architectures are compared based on the user requirements, heuristics and AHP. The values used in the AHP are in
