Smart Industry Services in Times of Internet of Things and Cloud Computing

Martin Serrano - Panos Dimitropoulos

Insight Centre for Data Analytics, NUI Galway, Galway City,Ireland
(Digital Enterprise Research Institute - DERI)
martin.serrano@deri.com

Sensap Microsystems, Athens City, Greece
pdimi@sensap.eu

1. Introduction - Internet of Things in the Manufacturing Industry

The benefits of the Internet of things (IoT) technologies in the area of manufacturing have motivated enormous progress and potentially are generating big economic impact. Based on the advent and deployment of RFID solutions, the Internet of things is being consolidated as the progress engine in the manufacturing sector and smart industry in general [Johnson02], [Rockwell04]. RFID deployments have exposed benefits associated with the reduction of labour and inventory costs, as well as other techno-economic benefits [Lee04], [Toffaletti10]. These benefits stem from the use of unique identification (including the ability for serialization), item level track and trace and enhanced track and trace, automated genealogy, elimination of the need for line-of-sight for data readability and, finally, historical tracing. This gave rise to a number of RFID deployments for manufacturing, which however tend to be isolated and focused on specific companies and cases studies [Brintrup08].

In general, RFID deployments in manufacturing cover all the different stages of the production process. For example, in the area of product design the EU FP6 PROMISE project [Promise04] has validated the RFID based linking of field usage data with the product design stage, with a view to improving future designs of products. In terms of production planning, RFID has been used to optimize production rescheduling [Hozak08], as well as dynamic improvements in production planning [Li06].

Several case studies have also focused on the production stage, mainly based on tracking and tracing of the production processes/steps towards improving quality [Huang07], scheduling and production decision making. Other (validated) RFID applications in manufacturing include storage management of perishable materials [Mills-Harris05], Internet-based inventory control [Zhou07], automating outbound shipments of a product after manufacturing [Wessel06], as well as reconfiguring machines in response to changed product configurations [Huang07]. Most of the above RFID-based solutions are custom system integrated on the basis of the specific manufacturing requirements (for various industries), and implemented in a way that data silos have been created rather than solutions derived from generalpurpose platforms using more large-deployed infrastructure (cloud).

IoT solutions for manufacturing have been gradually extended in order to include multiple sensors, actuators and devices of the shop floor in addition to RFID. Practical solutions have been developed as part of recent IoT projects (such as IoT@Work - see [Dürkop12] and [Gusmeroli12]), but also as part of IoT vendors' offering. Cisco, SAP and Bosch have undertaken prominent commercial efforts leading the market and opening a new vision towards how the Internet in general will look like in the future.

A prominent example is advertised by Ford Focus Electric, which has built its own Internet of Things that enables communication and data exchange across devices within its vehicles, but also between in-vehicle devices and the company that built it. Ford has built a cloud-based secure server enabling vehicle owners to access a wide range of information via an on-board wireless module and a smartphone app or through Ford's website. The vehicle information provided includes battery state of charge, overall efficiency, energy consumption, and braking regeneration. This infrastructure enables the issue of appropriate alerts in the case of problems. Furthermore, it provides the means for reporting the car's location when it's lost in a parking lot, being used by the owner's teenage drivers, or stolen1.

Cisco emphasizes on the convergence of factory systems with IT networks, as part of its wider portfolio of IoT-related solutions. On the other hand, SAP and Bosch promote the communication and interconnection of the numerous devices that comprise a plant for tasks such as manufacturing performance monitoring and predictive maintenance. Recently, solutions that combine IoT with the cloud (i.e., as promoted by OpenIoT) have been also reported [Soldatos12][Serrano13].

In general, IoT Cloud solutions are expected to play significant role in the manufacturing industry, as also proclaimed by the initiative Industry 4.02, a term introduced by representatives of German industry leaders, researchers, industry association, and unions.

2. Smarter Services by Service Composition in Cloud Environments

Currently it is more than evident the business benefits of cloud systems, apart of the reduction in maintenance cost the capacity to run more robust processes, cloud significantly increase systems flexibility to react to user service demands efficiently and by replacing, in a best practice manner, a plethora of proprietary software platforms with generic solutions supporting standardised development and scalable stacks over the Internet. Thus Cloud is ideally the best ecosystem for service composition. Research initiatives addressing this cloud-based design trend and inspired mainly by software oriented architectures (SOA) requirements argue that the future rely in application layers above virtual infrastructures that can meet various requirements whilst keeping a very simplistic, almost unmanaged network. IP for the underlying Internet for example, GENI NSF-funded initiative to rebuild the Internet [GENI, online Feb 2011] is an example of this. Others argue that the importance of wireless access networks requires a more fundamental re-design of the core Internet Protocols themselves [Clean Slate, Online April 2011][AKARI, Online May 2011]. Whilst this debate races nothing is a clear outcome in terms of information interoperability or data models sharing.

The service composition is a complex process; it implies the identification of service features and elements, as well as it implies the possible evaluation of operation and functionality before the new service can be composed. Thus it can be regulated by semantic rules where if multiple operations are required, then these operations are performed using the appropriate applications, as defined by service composition rules and/or polices defined by the data associations. Best practices in SOA suggest that a narrow focus on designing optimal networking protocols in isolation is too limited; instead a more abstracted view is required. This offers the advantage of non-dependency on physical infrastructures offering limited amount of services. In this view multiple services are now result of subservices, this method is commonly called composition. When meaning of various distributed protocols and delivering sub-services orchestrate multiple sub services, the operations (e.g., applications, computing processing, distribution of services, networking) can be done more efficiently. In other terms, a more realistic way of offering services is following mechanisms to organise operations according to changes in the parameters and based on users needs. However, realistically this new holistic view increasingly stops to become a matter of critical infrastructure, in this sense cloud computing infrastructures with virtualisation, as main driver is a promising alternative of solution to this stopping problem.

Figure 1. Service Composition processes representation on Cloud Environments

Figure 1 depicts the mentioned cloud service composition, its implementation relies on the inference plane [Serrano09], or knowledge layer where the exchange of information (Linked-Data structures) [Decker08] facilitates knowledge-driven support and generation of cloud composed services with operations by enabling interoperable information on networked connected objects [Hauswirth11]. From down to top and having cloud infrastructures representation as example, isolated components representations are depicted with no capacities of sharing information, linked data mechanisms are missing and "X" represented. In an upper Layer linked mechanism are represented and used to define...