Workshop "From Objects to Agents" (WOA 2019) Towards the Internet of Safe and Intelligent Postal+ Things Massimo Ancona Viviana Mascardi Nicoletta Noceti Francesca Odone DIBRIS, University of Genova, Genova, Italy name.surname@unige.it Waqas Ahsen Antonino Scribellito PostEurop, Brussels, Belgium europeanprojects@posteurop.org, antonino.scribellito@posteurop.org Abstract—SAFEPOST is an FP7 European project which infrastructure (vehicles, mailboxes, machines, letter carriers was active from April 2012 to July 2016 aimed at the “reuse etc.) for bringing data management to the next level. In and development of Security Knowledge assets for International almost the same years, the notion of Internet of Intelligent Postal supply chains”, as its full title explains. SAFEPOST addressed threats to postal security by designing and experiment- Things [5], [18] as a network of “intelligent devices that are ing a sensor network detection system including gas, radiation, capable of communicating with each other, making certain Raman spectroscopy and image-based sensors. In 2015, while decisions based on local information, and taking autonomous SAFEPOST was running, the US Postal Service and IBM and coordinated actions”, to quote [18], was born. In the suggested the idea of applying sensors to the postal infrastructure community working on agents and multiagent systems (MAS), components to bring the acquired data to the next supply chain level and optimize efficiency and costs, leading to an Internet of such intelligent devices would be named “agents”, and a Postal Things. Merging the SAFEPOST and Internet of Postal network consisting of them, would be named “MAS”. And in Things approaches and applying the result of their merge to fact, even if using a different terminology, the idea of making supply chains involving not only postal items, but also logistic IoT smarter and more intelligent by exploiting methodologies infrastructures and business processes, paves the way to an and approaches coming from the agent community, also started Internet of Safe Postal+ Things, IoSP+ T. The IoSP+ T can be further enriched and made smarter, more flexible, and intelligent, to flourish in those years [15], [26], [27], leading to many by adding agents below, inside, and on top of it. initiatives including the special session on “Internet of Things In this paper we provide our vision of the Internet of Safe and and Internet of Agents (ITIA)” at the IDC 2019 conference Intelligent Postal+ Things, IoSIP+ T, highlighting challenges and [30], and the explicit reference to IoT as a topic of interest in opportunities. the PRIMA 2019 call for papers [7]. Index Terms—Agents, Multiagent Systems, Internet of Intel- ligent Things, Internet of Postal Things, Safety, Supply Chain Despite intelligence on board of the interconnected devices, a wide adoption of Internet and IoT, with a consequent widespread dissemination of sensors, greatly amplifies secu- I. I NTRODUCTION rity and efficiency problems. In such a context, the expe- rience gathered in recent EU projects provides fundamental As observed by [19], [32], [33] among many others, Postal insights for future developments of e-logistic projects based Services are reporting mounting deficits every year all over on Machine-To-Machine (M2M) and IoT. On this respect we the world: in order to survive, they need to redesign their mention three recent EU e-logistics projects, SAFEPOST1 , business. In particular, first-class mail undergoes e-mail, sms, iCargo2 , and e-Freight3 which aimed at developing frame- chat and other forms of electronic replacements. “The worse works, reference models, and demonstrators to improve, via news is the Postal Service expects first-class mail volume to technologies at the state of the art, a safe and efficient continue dropping by nearly 50% over the next decade” [19]. transport of, respectively, postal items, cargo, and containers, If they want to survive, they must exploit the most recent throughout the EU. When in Europe these three projects were technological advances: sensors within the Internet of Things still active or just closed, in the US, USPS and IBM envisioned [6], Cloud Computing [51], Big Data technologies [73] and, the Internet of Postal Things. By generalizing USPS vision, above all, Artificial Intelligence. Based on these observations, we might consider the “Internet of Safe Postal+ Things” the US Postal Service (USPS) and IBM recently proposed to (IoSP+ T), where by “Postal+ ” we mean Postal items and take advantage of the dramatic decline of the cost of sensors and wireless data connectivity, and push the Internet of Things 1 https://www.posteurop.org/SAFEPOST, start: 01/04/2012; end: (IoT) into the postal domain. 30/07/2016. 2 https://cordis.europa.eu/project/rcn/100869/factsheet/en, start: 01/11/2011; In 2015, the USPS RARC Report [69] and Marsh and end: 30/04/2015. Piscioneri [49] presented the Internet of Postal Things (IoPT) 3 https://cordis.europa.eu/project/rcn/94475/factsheet/en, start: 01/01/2010; vision: applying sensors to the various component of the Postal end: 30/06/2013. 51 Workshop "From Objects to Agents" (WOA 2019) Fig. 1. The layered architecture of the SAFEPOST project. components together with more general logistic processes and (CPSS); (iii) the application layer, composed by the e- infrastructural architectures. And by exploiting agents, IoSP+ T logistic (service oriented) optimization code. could move towards the IoSIP+ T, namely, the “Internet of Safe • The treatment of security performed in each architectural and Intelligent Postal+ Things”. layer. SAFEPOST, iCargo, and e-Freight could naturally evolve The layered architecture shown in Figure 1 simplifies the into IoSIP+ T projects. development of certified code while providing high levels of In the IoSIP+ T context, safety and security are so closely assurance, and makes this process practical and affordable. It intertwined to be no longer considered two separate concerns: also simplifies code and artefact reuse to leverage investments. any system interfaced with the outside world has the potential A Postal Security Stamp (PSS) connects the SAFEPOST to expose security vulnerabilities. In particular, systems con- components together. It identifies postal items and associates nected to the Internet and the IoT need to be protected against place of origin, destination, content information, screening specialized targeted malware attacks and against a whole world history, image comparison data, track and trace mechanisms of hackers. This paper deals with safe and secure IoSIP+ T, with them. The information can be traced in real time by grounding its root into the authors’ experience in SAFEPOST authorized stakeholders. and in Artificial Intelligence in general, and agents and MAS SAFEPOST implementation is centered on a D2D (Door- in particular. After presenting the state of the art in postal To-Door) delivery mechanism completely controlled by human security - mainly represented by the results achieved by operators performing H2M (Human-To-Machine) & M2H SAFEPOST - in Section II, we analyse new research directions (Machine-To-Human) communication. Replacing (part of) a towards intelligent, dependable and resilient supply chains H2M mechanism by a direct M2M communication in an development with the adoption of agents together with IoT IoT environment, besides optimizing efficiency and costs of and cloud computing in Section III. Section IV is left to delivery, remarkably increases security problems. This is why conclusions and final remarks. SAFEPOST has been conceived to be secure by design: II. SAFEPOST AND THE S TATE OF THE A RT IN P OSTAL SAFEPOST implements a security level able to satisfy evolv- S ECURITY ing international regulations and standards while efficiently supporting the complexity of postal services market across Eu- “SAFEPOST: reuse and development of Security Knowl- rope, without increasing costs. Also, and more important for its edge assets for International Postal supply chains” is an FP7 implications in the development of an IoSIP+ T, SAFEPOST European project which spanned the period from April 2012 could be adopted as the minimum kernel in the development to July 2016, involving 20 partners for a total cost of nearly of future safe and dependable intelligent supply chain systems. 15 million euros. Its main design principles are: • The introduction of safety sensors, i.e. sensors expressly A. The Screening System and Safety Sensors designed for augmenting knowledge about safety and security information on parcels. The SAFEPOST Screening System is a hardware-software • A well defined hierarchical organization based on (i) infrastructure consisting of D-Tube, Raman and Radiak sen- the sensing layer, implemented by the Targeting and sors, and an image recognition system4 . Threat Handling component and devoted to the safety sensors management and data acquisition; (ii) the network 4 The information related to some sensors is confidential so just a short layer, implemented by the Common Postal Security Space overview of their specifications and functionalities can be published. 52 Workshop "From Objects to Agents" (WOA 2019) 1) The D-Tube: The D-tube is used for detecting explosives or narcotics in real time; the prototype developed within the SAFEPOST project uses a multi-element detector for the generation of the chemical profiles and is suitable for integration into the sorting facility conveyor belt flow. It operates in near real time with high precision and specificity by examining vapor substances. The system also takes into account the varying background vapors ubiquitously present in the air. This D-Tube system consists of three different sub- systems: • The Breathing Sub-system • The Air Supply Sub-system • The eNose Gas Sensor Sub-system Fig. 2. The image recognition system. The Breathing Sub-system is positioned over the conveyor belt. It uses a device to compress the packages in order to press out air from the inside of the packages. This procedure makes packages emit as much of the enclosed molecules as possible for further detection. This “active breathing” is key to make the system work consistently on a wide range of different packaging types. The Breathing System is designed to stay within the envelope of specifications for proper handling of the packages. The Air Supply System is a “sniffer” system positioned after the breathing device to “sniff” any molecules being emitted from the packages. It consists of several silicon hoses and acts as a “vacuum cleaner” to vacuum the packages on its side and Fig. 3. Package handling work-flow. on the top. This is being done just after the breathing system has compressed the package. The hoses in the “sniffer” system are then connected and lead the air flow to the Gas Sensor Sub- system, also called the eNose. The eNose chamber consists of B. The Targeting and Threat Handling Reasoning System 18 separate electronic sensors (“noses”) capable of detecting The Targeting & Threat Handling Reasoning System is a different molecules such as explosives and narcotics. high level decision support system which combines informa- 2) Radiak and Raman Sensors: The radiation sensor is tion from both the SAFEPOST sensors and other sources to based on a semiconductor detector of High Purity germanium conduct risk assessments. Risk assessment is driven by the (HPGe detector) which measures ionizing radiation by means work-flow of the package handling represented, in a simplified of the number of charge carriers set free in the detector way, in Figure 3. Information gathered from sensors and from material, which is arranged between two electrodes, by the the stakeholders involved in the process is used to assess radiation. Ionizing radiation produces free electrons and holes. the risk that a parcel is a potential threat throughout the Under the influence of an electric field, electrons and holes entire D2D postal delivery supply chain. The system discovers travel to the electrodes, where they result in a pulse that can potential threats in real-time and assesses those not detectable be measured in an outer circuit. by physical screening. When a threat is detected, the system 3) Image Recognition System: The role of the Image helps users decide on which plan to follow based upon the Recognition System in SAFEPOST (see a sketch in Figure threat, time and resources available. The Targeting and Treat 2) is to allow postal operators to screen the exterior of the Handling Reasoning System operates at the intermediate level parcels in order to detect damages and signs of tampering and performs different fusion services onto data in the CPSS [58]. The Image Recognition System provides information to to address different phases of the risk management process as the CPSS as well as to the human postal operator. Information follows: sent to the CPSS can be integrated with the results of other • A risk assessment sub-system fuses all available informa- sensors or with previous scans of the same parcel in order to tion regarding letters and parcels to determine which parts compare if and how the parcel changed over time and better of the sorting facility convey or belts should be subject evaluate its tampering risk. At the same time, as the system to additional screening on top of the mandated ones. This performs the parcel analysis on the fly, human operators can step of the process uses available information from, e.g., be immediately informed of suspicious parcel’s features and alert levels set by security agencies, intelligence data- can intervene on the parcel, according to the risk management bases, information from cargo operators and customs, be- policies implemented by the postal operator. sides the information coming from SAFEPOST sensors. 53 Workshop "From Objects to Agents" (WOA 2019) • The results of the detection are combined with other E. Related Work in Postal Security available data to determine what should be done with the suspicious parcels. In particular, using information about After the Yemen bomb plot in 20106 , the postal security the uncertainties in the detection result and combining management entered its biggest reform in modern times. In these with other available information enables the system 2012 the Universal Postal Union (UPU), the United Nations to choose different handling strategies (with associated (UN) organization coordinating global postal policies, issued different delays in cycle time) for different risks. two postal security standards binding all of its 192 members countries: • S58, Postal Security Standards - General Security Mea- sures which defines the minimum physical and process security requirements available to critical facilities within the postal network; • S59, Postal Security Standards - Office of Exchange and International Airmail Security which defines minimum requirements for security operations relating to the air transport of international mail. Besides regional legislation in the transportation sector, which have been rapidly evolving as well over the past few years with major implications on postal security7 , also academic literature on security in the supply chain has become huge after September 11, 2001: as observed by Männistö in [50], “The terrorist attacks of September 11, 2001 raised major concerns about the vulnerability of global transportation sys- tems to transnational crime and terrorism. Although the attacks Fig. 4. The Common Postal Security Space. occurred in the context of passenger transport, they spurred unprecedented academic research on supply chain security (SCS)”. Männistö defines a supply chain crime taxonomy, carries out a deep literature review showing that the SCS C. The Common Postal Security Space (CPSS) discipline is more empirically grounded and diverse than the The CPSS is shown in Figure 4 and is based on the PostEu- previous literature reviews suggest, and discusses a case study rop framework5 as operating model, which aligns EU Policy in the international postal service from the Swiss perspective. and Legislation with IT security management and industry To the best of our knowledge, that work is one of the wide best practice processes, through a shared network. The more recent and complete academic documents dealing with CPSS is integrated with the screening systems, target and safety and security in the supply chain in general, and in threat handling reasoning and information sharing via the the postal sector in particular. Compared to SAFEPOST8 it universally trusted Postal Security Stamp. complements its practical results with a strong theoretical underpinning by providing a taxonomy which can serve as a D. The Optimization Component unifying framework in the supply chain crime area, and which can be generalized and adapted to other domains. Although This component consists of a real-time logistics optimiza- Männistö’s work also includes a practical component, leading tion system that automatically reads GPS tracking data and to concrete suggestions to the Swiss Post and Swiss authorities updates the logistics plan accordingly. This is more than based on the results of the case analysis, it cannot compete simply updating the Estimated Time of Arrival (ETA) of a with the practical solutions to threat handling proposed within vehicle, as the system automatically fixes resulting logistics such a large project as SAFEPOST. problems using intelligent optimization algorithms. For exam- Given that the SAFEPOST approach is fully compliant with ple, if a vehicle is delayed so that it will not arrive at its the most recent regional and UPU standards, and that the destination in time to undertake the next planned work, the academic literature we are aware of in the postal safety and optimization component automatically reassigns the work to security areas is connected and complementary to the project, the next best resource available. The real-time rescheduling we can assess that SAFEPOST findings are the state of the art means that the logistics plan is constantly updated (working in postal security. within operational constraints and rules) so that at any point in time the system “knows” what should be happening next, 6 https://en.wikipedia.org/wiki/Cargo planes bomb plot even if this is now different from the original plan at the start 7 https://ec.europa.eu/transport/modes/air/security/legislation en;https: of the day. //www.faa.gov/regulations policies/faa regulations/ 8 Männistö’s Ph. D. Thesis is not fully disjoint from SAFEPOST: he received 5 http://www.posteurop.org/VisionandMission funding from SAFEPOST, as stated in his thesis acknowledgments. 54 Workshop "From Objects to Agents" (WOA 2019) III. B EYOND SAFEPOST: TOWARDS AN I O SIP+ T for obtaining an immediate reaction to unpredictable events through smart organizations: both resiliency and dependability As discussed in Section II, SAFEPOST is a layered, dis- imply safety and security requiring advanced forms of context- tributed architecture with sophisticated sensors at the lowest awareness. The meaning of context depends on the application level, organizations that must protect their data and privacy on domain and may involve many aspects, including location, the one hand, and must collaborate to reach a safer manage- time of day, emotional state of the user, orientation, and even ment of postal items on the other, complex interactions among the preferences or identities of people within an environment. the parties involved, reasoning and optimization systems at In an IoSIP+ T, the context is also represented by data coming the top of the architecture. The holonic MAS metaphor [64] from advanced physical sensors. In order to be useful, sensory is extremely suitable to describe SAFEPOST: some sensors data need to be located, described, measured, and analysed on like the image recognition system depicted in Figure 2 are the fly in real-time, which requires sophisticated approaches to so sophisticated, that may be seen as MASs themselves. The sensor fusion. SAFEPOST already provides a first answer to same holds for the different providers of postal services, which the need of dependability, resilience and context awareness; in could be seen as individual agents if we analyze the high level its current setting, context- and self-awareness are supported interactions among them, taking place within the CPSS, but by the specific safety sensors discussed in Section II-A whose can also be seen as MASs due to the many components they outputs are fused within the targeting and threat handling consist of, and their complex dynamics. reasoning system discussed in Section II-B. More intelligence can be added to SAFEPOST’s architec- In a more general IoSIP+ T perspective, SAFEPOST could ture, both at the sensor level – by making sensors smarter –, be extended to integrate other sensors into a single MAS made and at the global architecture level – by making it more flexible secure by design [31] and supporting different levels of aware- and adaptable, and by improving the existing optimization ness operated in an IoT, within a Cloud environment. In the component, which could be based on agents [8], [29], [61], following we discuss approaches, methods and techniques that [70]. The reasoning component could take advantage of agents could be integrated within the existing SAFEPOST framework, as well, along the lines of some recent proposals for agent- to transform it into an IoSIP+ T. based distributed reasoning including [9], [38], [54]. But the “SAFEPOST of the future”, besides intelligent, must A. CPS, Spimes and Smart Objects also be resilient. Resilience is the capacity to quickly recover from difficul- SAFEPOST safety sensors operate in real-time and are de- ties; in a supply chain context, it can be seen as the ability signed with the dependable and resilient architecture of fully- to react in a timely fashion to unexpected external events protected Cyber Physical System (CPS). Safety and security including delayed delivery, reduced exchange of information are design dimensions of a CPS: safety is aimed at protecting with other companies in the chain, hardware/software failures, the systems from accidental events while security is limited to but also more disruptive ones such as floods, earthquakes, hostile actions, and both share the goal of protecting a CPS acts of terrorism. Dependability of a system reflects the user’s from failures. As pointed out in [62] when safety and security degree of trust in that system: it measures the numerical are aligned, namely when actions performed for enforcing extent of the user’s confidence that the system will operate safety do not contrast with actions performed for enforcing as expected. Implementing dependable and resilient supply security, they make the enclosing CPS almost inviolable. A chains is a strategic choice for mitigating the risks [35]. The theoretical concept which has recently emerged as an evolution notion of dependability9 , together with responsiveness, agility, of CPS and sensors in the direction of real time and context- cost, and assets, is one of the five Standard Strategic Metrics aware tracking, is that of the spime11 a uniquely identifiable of the Supply Chain Operations Reference model, SCOR10 . object whose real-time attributes can be continuously tracked. Delivery dependability, also known as delivery reliability [63] Smart objects [39], namely computers equipped with sensors is the ability, for supply chains,“to exactly meet quoted or and/or actuators, and a communication device, are examples anticipated delivery dates and quantities” [72]. Specific sensors of spimes. They are described as the building blocks for and relative fusion algorithms can help extracting several kinds the IoT [43] and can be embedded in cars, light switches, of contextual data, raising the system context awareness which thermometers, billboards, or machinery. The gap between a represents a key ingredient of the IoT. Quoting [53], “...smart smart object and an intelligent software agent is very small: the connectivity with existing networks and context-aware com- smarter the object, the closer to an agent [60]. The integration putation using network resources is an indispensable part of of smart objects in SAFEPOST would definitely increase not IoT. With the growing presence of WiFi and 4G-LTE wireless only its context awareness, but also its ability to self-adapt and Internet access, the evolution towards ubiquitous information self-repair thanks to the smart objects actuators which can play and communication networks is already evident”. Today’s an active role, differently from traditional, passive sensors. The companies are already adopting resilience and dependability ability to manage itself is a “must” for the SAFEPOST of the future, given that SAFEPOST should be able to survive 9 Named “reliability” and meaning “Perfect Order Fulfillment”. 10 http://www.apics.org/apics-for-business/products-and-services/ 11 Spime is a neologism introduced by B. Sterling (2005) as the contraction apics-scc-frameworks/scor of space and time. 55 Workshop "From Objects to Agents" (WOA 2019) to disasters, to continue keeping the infrastructure as safe as to increase its dependability, in particular when boosted by possible. The opportunities to increase SAFEPOST reliability agents [46]. A challenging research direction is “real-time thanks to self* approaches are huge: self-adaptive and self- weak signal detection mining”, and its integration with the organizing MASs [1], [28], [55], [65] are very close to results obtained by a more traditional mining of IoT data. autonomic computing systems, namely systems that “manage In fact, a really safe logistic system, should also take weak themselves according to an administrators goals. New compo- signals of threats coming from news, social networks, the nents integrate as effortlessly as a new cell establishes itself web, into account. As observed in [14], “Lone wolf terrorists in the human body” [41]. An “autonomic SAFEPOST” would pose a large threat to modern society. The current ability to give many advantages, but would also raise many challenges. identify and stop these kinds of terrorists before they commit In particular, autonomy in self-adapting, self-organizing and a terror act is limited since they are hard to detect using self-repairing should balance compliance to existing technical, traditional methods. However, these individuals often make use legal, and even ethical constraints. To reach this goal, designers of Internet to spread their beliefs and opinions, and to obtain should be able to verify in advance (at design time, when information and knowledge to plan an attack. Therefore there the system is still under test) as many properties of the is a good possibility that they leave digital traces in the form “autonomic SAFEPOST” as possible via traditional techniques of weak signals that can be gathered, fused, and analyzed”. based on model-checking. At runtime, when the system has An agent-based approach might turn useful to tackle the last been deployed, it should undergo a continuous monitoring to tasks, as discussed for example in [57], [76]. early detect those anomalies that static verification techniques could not capture, and report them to a human controller. C. Integrating IoT and Cloud Computing for Logistics While these problems are far to be solved, in particular for Although IoT and cloud computing are different paradigms, large and complex systems as an “autonomic SAFEPOST” they play complementary roles in tackling emerging needs of would be, some results achieved within the MAS community the current world, and both are recognized as being extremely seem to go in the right direction. Model checking MASs suitable to supply chain management. While the IoT mainly has a long tradition which dates back to the beginning of consists of device connections via the Internet, the role of the millenium [11], [12], [47], [74] and, although not scaling the cloud is to deliver data, applications, streams, images, well to large systems, could be used for the design-time static and other digital objects in a distributed context. The huge verfication. Recently, runtime verification mechanisms suitable implications of IoT for logistics have been explored by Cisco for MASs in particular, and for distributed system in general, and DHL [48] among the others and are raising more and have been proposed [2], [4], [23]. Attempts to integrate static more attention [40], [52], [66], mainly when combined with and runtime verification techniques in the MAS domain are big data management [42]. A similar or maybe even greater discussed in [3] and in [24], [25]: while the first work interest can be observed for the adoption of cloud computing deals with the relationships between Linear Temporal Logic for logistics and supply chain management, which - besides and the trace expression formalism used for MAS runtime many scientific and popular articles (see for instance the recent verification, the last ones address the problem of validating an works [16], [67]) - also lead to patents [36]. The integration of abstract environment designed for model checking purposes, IoT with cloud computing [13] offers an additional potential at runtime. benefit to significantly reduce costs on a pay-per-use base and with an improved customer service based on rationalization B. Real-Time Data Mining of operations, through optimization and increased operating A real-time management of threats, like in SAFEPOST’s and economic efficiencies and supporting the development of Targeting and Threat Handling Reasoning System, requires new services and business models. Several postal and logistic a real-time analysis of the acquired data and new forms of companies are already offering their service from the cloud dependable data mining algorithms. As observed in [68], “data and the combination of IoT with intelligent CPSs makes them mining has typically been applied to non-real-time analyti- accessible anytime and anywhere. This combination will be a cal applications. Many applications, especially for counter- key enabler of the IoSIP+ T, and agents will play a relevant role terrorism and national security, need to handle real-time to add intelligence, security, and flexibility to it, as discussed threats. [...] This means that the data mining algorithms have for example in [45], [75], [77]. to have the ability to recover from faults as well as main- tain security, and meet real-time constraints all in the same IV. C ONCLUSIONS AND F UTURE W ORK program.” Real time data mining is a hot research topic, as In its most recent mobility report, Ericsson forecasts that witnessed by many existing tools [10], [56], papers [22], [37], around 29 billion connected devices will be available by 2022, [44], [71], and even patents [20]. In the same way as massive of which around 18 billion related to IoT. Connected IoT data generated by the IoT can be analysed and managed with devices will include cars, machines, meters, sensors, point- data mining techniques adapted to cope with big data and of-sales terminals, consumer electronics and wearable [34]. data streams [17], existing data mining algorithms operating Between 2016 and 2022, IoT devices are expected to increase in real-time could be adopted to manage data generated by at a compound annual growth rate of 21 percent, driven by new an IoSIP+ T system, in order to cope with its complexity and use cases. In [21] the author discusses seven ways the IoT will 56 Workshop "From Objects to Agents" (WOA 2019) change our lives. Besides boosting remote work and increasing [9] I. Benelallam, Z. Erraji, G. E. Khattabi, and E. H. Bouyakhf. Dynamic speed, accessibility, efficiency, and productivity, he points out JChoc: A distributed constraints reasoning platform for dynamically changing environments. In International Conference on Agents and that IoT will transform how companies track and manage Artificial Intelligence, pages 20–36. Springer, 2015. their inventory, since smart devices will be able to keep tabs [10] A. Bifet, G. Holmes, R. Kirkby, and B. Pfahringer. MOA: Massive on inventory changes completely automatically, moving from Online Analysis. J. Mach. Learn. Res., 11:1601–1604, Aug. 2010. [11] R. H. Bordini, M. Fisher, C. Pardavila, W. Visser, and M. Wooldridge. “smart homes” to “smart offices” and “smart warehouses”. Model checking multi-agent programs with CASP. In International According to [59], companies in the 2016 Fortune 500 list12 Conference on Computer Aided Verification, pages 110–113. Springer, are implementing real-time resilient and dependable supply 2003. [12] R. H. Bordini, M. Fisher, W. Visser, and M. Wooldridge. Verifying multi- chain IoT platforms able to solve key questions pertinent agent programs by model checking. Autonomous agents and multi-agent shipment’s context-awareness, like continuous knowledge of systems, 12(2):239–256, 2006. its space-temporal position and precise forecasting of delivery [13] A. Botta, W. de Donato, V. Persico, and A. Pescapé. On the integration time. Many companies are already boosting their business of cloud computing and Internet of Things. In Proceedings of the 2014 International Conference on Future Internet of Things and Cloud, thanks to IoT and/or cloud computing, and the others will FICLOUD ’14, pages 23–30, Washington, DC, USA, 2014. IEEE follow. For those working in the logistic sector, moving to an Computer Society. IoSIP+ T would require investments in new intelligent safety [14] J. Brynielsson, A. Horndahl, F. Johansson, L. Kaati, C. Mårtenson, and P. Svenson. Harvesting and analysis of weak signals for detecting lone sensors, an infrastructure supporting not only communication wolf terrorists. Security Informatics, 2(1):11, Jul 2013. among these sensors, but also the possibility to verify their [15] D. Calvaresi, M. Marinoni, A. Sturm, M. Schumacher, and G. Buttazzo. behaviour (including communicative behaviour) at runtime, a The challenge of real-time multi-agent systems for enabling IoT and CPS. In Proceedings of the international conference on web intelligence, shift in the business process to include some “security stamp” pages 356–364. ACM, 2017. à la SAFEPOST, attached to physical objects, and an injection [16] R. Cao, D. G. Schniederjans, and M. Schniederjans. Establishing the of intelligence in all of those processes which are not routinary, use of cloud computing in supply chain management. Operations Management Research, 10, 02 2017. and require to perform some reasoning or other sophisticated [17] F. Chen, P. Deng, J. Wan, D. Zhang, A. V. Vasilakos, and X. Rong. Data tasks. Moving to this new setting would be worth the costs, mining for the internet of things: Literature review and challenges. Int. since the integration of security sensors in CPSs connected J. Distrib. Sen. Netw., 2015:12:12–12:12, Jan. 2015. [18] Y. Chen and H. Hu. Internet of intelligent things and robot as a service. in an IoT network as smart objects (or agents) and operating Simulation Modelling Practice and Theory, 34:159 – 171, 2013. in a cloud environment in real time seems one of the most [19] M. A. Cusumano. Can services and platform thinking help the U.S. promising approaches for realizing dependable, resilient, and Postal Service? Commun. ACM, 55(4):21–23, Apr. 2012. intelligent supply chains. [20] K. Dasgupta, B. Nilanjan, C. Dipanjan, M. Sumit, N. Seema, J. Anupam, and R. Angshu. Real-time data mining, 2017. [21] J. DeMers. 7 ways the Internet of Things will change businesses in R EFERENCES 2017, 2017. [22] W. Fan and A. Bifet. Mining big data: Current status, and forecast to the future. SIGKDD Explor. Newsl., 14(2):1–5, Apr. 2013. [1] D. Ancona, D. Briola, A. Ferrando, and V. Mascardi. Global protocols as first class entities for self-adaptive agents. In AAMAS, pages 1019–1029. [23] A. Ferrando, D. Ancona, and V. Mascardi. Decentralizing MAS ACM, 2015. monitoring with DecAMon. In AAMAS, pages 239–248. ACM, 2017. [2] D. Ancona, S. Drossopoulou, and V. Mascardi. Automatic generation [24] A. Ferrando, L. A. Dennis, D. Ancona, M. Fisher, and V. Mascardi. of self-monitoring MASs from multiparty global session types in Jason. Recognising assumption violations in autonomous systems verification. In DALT, volume 7784 of Lecture Notes in Computer Science, pages In AAMAS, pages 1933–1935. International Foundation for Autonomous 76–95. Springer, 2012. Agents and Multiagent Systems Richland, SC, USA / ACM, 2018. [3] D. Ancona, A. Ferrando, and V. Mascardi. Comparing trace expressions [25] A. Ferrando, L. A. Dennis, D. Ancona, M. Fisher, and V. Mascardi. and Linear Temporal Logic for runtime verification. In Theory and Verifying and validating autonomous systems: Towards an integrated Practice of Formal Methods, volume 9660 of Lecture Notes in Computer approach. In RV, volume 11237 of Lecture Notes in Computer Science, Science, pages 47–64. Springer, 2016. pages 263–281. Springer, 2018. [4] D. Ancona, A. Ferrando, and V. Mascardi. Parametric runtime veri- [26] G. Fortino. Agents meet the IoT: Toward ecosystems of networked smart fication of multiagent systems. In AAMAS, pages 1457–1459. ACM, objects. IEEE Systems, Man, and Cybernetics Magazine, 2(2):43–47, 2017. 2016. [5] A. Arsénio, H. Serra, R. Francisco, F. Nabais, J. Andrade, and E. Ser- [27] G. Fortino, A. Guerrieri, W. Russo, and C. Savaglio. Integration of rano. Internet of intelligent things: Bringing artificial intelligence agent-based and cloud computing for the smart objects-oriented IoT. into things and communication networks. Inter-cooperative Collective In Proceedings of the 2014 IEEE 18th international conference on Intelligence: Techniques and Applications, page 1, 2014. computer supported cooperative work in design (CSCWD), pages 493– 498. IEEE, 2014. [6] L. Atzori, A. Iera, and G. Morabito. The Internet of Things: A survey. Computer networks, 54(15):2787–2805, 2010. [28] L. Gardelli, M. Viroli, M. Casadei, and A. Omicini. Designing self- organising environments with agents and artefacts: a simulation-driven [7] M. Baldoni, Y. Sakurai, M. Dastani, B. Liao, and R. Zalila-Wenkstern. approach. IJAOSE, 2(2):171–195, 2008. The 22nd international conference on principles and practice of multi- agent systems (PRIMA), 2019. [29] J. Gjerdrum, N. Shah, and L. G. Papageorgiou. A combined optimization and agent-based approach to supply chain modelling and performance [8] M. Barbati, G. Bruno, and A. Genovese. Applications of agent-based assessment. Production Planning & Control, 12(1):81–88, 2001. models for optimization problems: A literature review. Expert Systems with Applications, 39(5):6020–6028, 2012. [30] V. Gorodetky, V. Maric, and P. Skobelev. Internet of things and internet of agents (ITIA): special session at idc 2019, 2019. [31] J. Granjal, E. Monteiro, and J. S. Silva. Security for the internet of 12 The Fortune 500, http://fortune.com/fortune500/, is an annual list com- things: a survey of existing protocols and open research issues. IEEE piled and published by Fortune magazine that ranks 500 of the largest United Communications Surveys & Tutorials, 17(3):1294–1312, 2015. States corporations by total revenue for their respective fiscal years. The list [32] F. Guerrini. Achieving high performance in the post and parcel industry, includes publicly held companies, along with privately held companies for Accenture research and insights. https://www.accenture.com/us-en/ which revenues are publicly available. insight-achieving-high-performance-post-and-parcel-industry, 2015. 57 Workshop "From Objects to Agents" (WOA 2019) [33] F. Guerrini. How big data and the Internet Of Things will change the [57] G. Neubert, Y. Ouzrout, and A. Bouras. Collaboration and integration postal service. http://www.forbes.com/sites/federicoguerrini/2014/07/03/ through information technologies in supply chains. arXiv preprint how-big-data-and-the-internet-of-things-will-change-the-postal-service, arXiv:1811.01688, 2018. 2015. [59] D. Palmquist and T. Leal. The IoT supply chain: 3 things you should [34] N. Heuveldop. Ericsson mobility report, 2017. know, 2016. [35] W. Ho, T. Zheng, H. Yildiz, and S. Talluri. Supply chain risk [60] P. Pico-Valencia and J. A. Holgado-Terriza. Agentification of the Internet management: a literature review. International Journal of Production of Things: A systematic literature review. International Journal of Research, 53(16):5031–5069, 2015. Distributed Sensor Networks, 14(10):1550147718805945, 2018. [36] P. Jaeger and R. Lindenlaub. Cloud logistics, 2016. [61] J. Rouzafzoon and P. Helo. Developing logistics and supply chain man- [37] D. Jankov, S. Sikdar, R. Mukherjee, K. Teymourian, and C. Jermaine. agement by using agent-based simulation. In 2018 First International Real-time high performance anomaly detection over data streams: Grand Conference on Artificial Intelligence for Industries (AI4I), pages 35–38. challenge. In Proceedings of the 11th ACM International Conference on IEEE, 2018. Distributed and Event-based Systems, DEBS ’17, pages 292–297, New York, NY, USA, 2017. ACM. [62] G. Sabaliauskaite and A. P. Mathur. Aligning cyber-physical system [38] A. Jarraya, A. Bouzeghoub, A. Borgi, and K. Arour. Distributed safety and security. In Complex Systems Design & Management Asia, collaborative reasoning for HAR in smart homes. In Proceedings of the pages 41–53. Springer, 2015. 17th International Conference on Autonomous Agents and MultiAgent [63] R. Sarmiento, M. Byrne, L. Rene Contreras, and N. Rich. Delivery Systems, pages 1971–1973. International Foundation for Autonomous reliability, manufacturing capabilities and new models of manufactur- Agents and Multiagent Systems, 2018. ing efficiency. Journal of Manufacturing Technology Management, [39] M. Kallman and D. Thalmann. Modeling Objects for Interaction Tasks. 18(4):367–386, 2007. Springer, 1998. [64] M. Schillo and K. Fischer. Holonic multiagent systems. Manufacturing [40] B. Karakostas. Towards Autonomous IoT Logistics Objects. IGI Global, Systems, 8(13):538–550, 2002. 2017. [65] G. D. M. Serugendo, M.-P. Gleizes, and A. Karageorgos. Self- [41] J. O. Kephart and D. M. Chess. The vision of autonomic computing. organization in multi-agent systems. The Knowledge Engineering IEEE Computer, 36(1):41–50, 2003. Review, 20(2):165–189, 2005. [42] N.-H. Kim. Design and implementation of Hadoop platform for process- [66] J.-P. Su, C.-A. Wang, Y.-C. Mo, Y.-X. Zeng, W.-J. Chang, L.-B. Chen, ing big data of logistics which is based on IoT. International Journal D.-H. Lee, and C.-H. Chuang. i-Logistics: An intelligent logistics system of Services Technology and Management, 23(1-2):131–153, 2017. based on Internet of Things. In 2017 International Conference on [43] G. Kortuem, K. Fahim, S. Vasughi, and F. Daniel. Smart objects as Applied System Innovation (ICASI), 05 2017. building blocks for the Internet of Things. IEEE Internet Computing, 14(1):44–51, 2010. [67] N. Subramanian and M. D. Abdulrahman. Logistics and cloud com- [44] K. Kwon, D. Kang, Y. Yoon, J.-S. Sohn, and I.-J. Chung. A real time puting service providers? Cooperation: a resilience perspective. In process management system using RFID data mining. Computers in Production Planning and Control, 2017. Industry, 65:721–732, 05 2014. [68] B. Thuraisingham, C. C. Latifur Khan, J. Maurer, , and M. Ceruti. [45] P. Leitao, S. Karnouskos, L. Ribeiro, J. Lee, T. Strasser, and A. W. Dependable real-time data mining. In Proceedings of the Eighth IEEE Colombo. Smart agents in industrial cyber–physical systems. Proceed- International Symposium on Object-Oriented Real-Time Distributed ings of the IEEE, 104(5):1086–1101, 2016. Computing, pages 158–165, 2005. [46] G. Lombardo, P. Fornacciari, M. Mordonini, M. Tomaiuolo, and [69] USPS. The Internet of Postal Things. Report Number RARC-WP-15- A. Poggi. A multi-agent architecture for data analysis. Future Internet, 013 https://www.uspsoig.gov/sites/default/files/document-library-files/ 11(2):49, 2019. 2015/rarc-wp-15-013 0.pdf, 2015. [47] A. Lomuscio, H. Qu, and F. Raimondi. MCMAS: A model checker for [70] R. R. van Lon, J. Branke, and T. Holvoet. Optimizing agents with genetic the verification of multi-agent systems. In International conference on programming: an evaluation of hyper-heuristics in dynamic real-time computer aided verification, pages 682–688. Springer, 2009. logistics. Genetic programming and evolvable machines, 19(1-2):93– [48] J. Macaulay, L. Buckalew, and C. Gina. Internet of things in logistics – 120, 2018. a collaborative report by DHL and Cisco on implications and use cases [71] M. Vanni, S. E. Kase, S. Karunasekara, L. Falzon, and A. Harwood. for the logistics industry, 2015. RAPID: real-time analytics platform for interactive data-mining in a [49] B. Marsh and P. Piscioneri. The Internet of Postal Things. In 2015 decision support scenario, 2017. International Conference on Collaboration Technologies and Systems (CTS), pages 3–4, June 2015. [72] S. K. Vickery, C. Dröge, and R. E. Markland. Dimensions of man- [50] B. Marsh and P. Piscioneri. The internet of postal things, collaboration ufacturing strength in the furniture industry. Journal of Operations technologies and systems conference, 2015. Management, 15(4):317–330, 1997. [51] P. Mell and T. Grance. The NIST definition of cloud computing. http:// [73] S. J. Walker. Big data: A revolution that will transform how we live, nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-145.pdf, work, and think. International Journal of Advertising, 33(1):181–183, 2011. 2014. [52] A. Meola. How IoT logistics will revolutionize supply chain manage- [74] M. Wooldridge, M. Fisher, M.-P. Huget, and S. Parsons. Model ment, 2016. checking multi-agent systems with MABLE. In Proceedings of the first [53] S. Mishra. Design and evaluation of wireless and IoT based healthcare international joint conference on Autonomous agents and multiagent system. International Journal of Current Engineering and Technology, systems: part 2, pages 952–959. ACM, 2002. 7(2), 2017. [75] G. Yıldırım and Y. Tatar. Simplified agent-based resource sharing [54] A.-W. Mohammed, Y. Xu, M. Liu, and H. Hu. Semantical markov logic approach for WSN-WSN interaction in IoT/CPS projects. IEEE Access, network for distributed reasoning in cyber-physical systems. Journal of 6:78077–78091, 2018. Sensors, 2017, 2017. [55] S. Monica, F. Bergenti, M. B. Blake, G. Cabri, and U. Wajid. Adaptive [76] Y. Zhang, G. Zhang, J. Wang, S. Sun, S. Si, and T. Yang. Real- computing (and agents) for enhanced collaboration (ACEC 2018). In time information capturing and integration framework of the internet WETICE, pages 1–2. IEEE Computer Society, 2018. of manufacturing things. International Journal of Computer Integrated [56] G. D. F. Morales. SAMOA: a platform for mining big data streams. In Manufacturing, 28(8):811–822, 2015. WWW, 2013. [77] S. Ziegler, A. Skarmeta, J. Bernal, E. E. Kim, and S. Bianchi. ANASTA- [58] N. Noceti, L. Zini, and F. Odone. A multi-camera system for damage and CIA: Advanced networked agents for security and trust assessment in tampering detection in a postal security framework. EURASIP Journal CPS IoT architectures. In 2017 Global Internet of Things Summit on Image and Video Processing, 2018(1):11, 2018. (GIoTS), pages 1–6. IEEE, 2017. 58