CONCEPT HIERARCHIES FOR SENSOR DATA FUSION IN THE COGNITIVE IoT Franco Cicirelli and Giandomenico Spezzano Institute for High Performance Computing and Networking (ICAR) CNR - National Research Council of Italy 87036 Rende(CS) - Italy Email: cicirelli@icar.cnr.it, spezzano@icar.cnr.it KEYWORDS Sensor data fusion; Internet of Things; Multi-agent sys- tems; Statecharts; MAB museum. ABSTRACT Sensor data fusion refers to technological solutions aiming at collecting, classifying and complementing data coming from multiple sensors. It has the potential of enabling context awareness which, on the other hand, represents a huge potential to be exploited in the field of IoT applications. Sensor fusion and IoT have to deal with multi-faced issues like heterogeneity, sensor/actuator management, data accuracy and reliability. This paper proposes a multi-tier approach dealing with sensor fusion and IoT aspects in a modular way. The approach relies on the use of the agent metaphor, statecharts and on the Rainbow multi-agent platform. Agents can be dynamically added and removed from an application thus promoting system openness and scalability. Heterogeneity and distribution issues are trans- parently managed by Rainbow which hides the physical layer on top of which the applications are built. As a significant case study, the approach was exploited for the implementation of a working prototype devoted to improve security of some artworks (statues) of the MAB museum located in the city of Cosenza, Italy. INTRODUCTION Nowadays, sensors are exploited in a huge variety of applications ranging from healthcare, transportation and lo- gistic, surveillance, environmental monitoring, smart city and so forth. The sensing capabilities of the infrastructures and devices surrounding our daily lives are constantly improving and becoming more affordable. The widespread diffusion of sensors was also favored by the ever-increasing attention which is given to the field of the so called “Internet of Things” (IoT) (Miorandi et al., 2012). The basic idea of the IoT is a pervasive presence around us of a variety of things or objects such as RFID tags, sensors, actuators, mobile phones, etc. which, through unique addressing schemes, are able to interact with each other and cooperate with their neighbors to reach common goals (Atzori et al., 2010). Important issues which are related to the development of significant sensor-based applications are data collection, clas- sification and complementation. Sensor data fusion, or simply sensor fusion (SF) (Karimi, 2013), refers to technological solutions having the aim of addressing the above issues. The goal is to increase both the accuracy and reliability of sensed data as well as to enable context awareness (Bicocchi et al., 2014; Karimi, 2013; Schilit et al., 1994). Context awareness refers to the capability of disclosing information about the context (or situation) in which the data get acquired/generated. This allows to validate an event or an assumption made on sensed data, or to compensate the lack of complete information about the sensed environment thus permitting to take decisions and/or properly react to complex environmental stimuli. As an example, a person may not see flames under the hood of a car, but the smell of burning rubber and the heat coming from the dash would suggest that it is prudent to leave the car because the engine is on fire. This paper proposes an approach for the development of distributed SF applications based on the IoT paradigm. The approach promotes separation of concerns through the use of a multi-tier architecture which allows to deal with SF and IoT issues in a modular and orthogonal way. The agent metaphor (Woolridge and Wooldridge, 2001) is used to structure the application logic, whereas agents’ behavior is modeled by using statecharts (Booch et al., 2000; Cicirelli et al., 2011; Harel, 1987; Kielar et al., 2014). Statecharts are a state- based formalism which allows to specify complex and time- dependent behaviors by using a graphical notation. Complexity of a model is dealt with the use of hierarchical constructs. Proposed approach permits the exploitation of concept hi- erarchies which allow the classification of high-level situation as well as the definition of reaction-driven policy which work on a multiplicity of low-level events. Concept hierarchies are useful to represent a system as a set of abstractions normally used by humans when reasoning about complex systems. The concepts which are introduced in different tiers, or within the same tier, can be related together (i.e., orchestrated) in order to specify complex reaction patterns to the stimuli coming from the external environment. The overall goal is to move from the IoT towards the Cognitive IoT (Tsai et al., 2014; Wu et al., 2014) where objects interact and operate by acquiring knowledge from the surrounding environment and by following a context-aware perception-action operational cycle. Each tier, in the proposed architecture, groups agents on the basis of some predefined agent roles. For instance, the role awareness was defined, and the corresponding tier will contain agents devoted to managing the knowledge acquired on the whole system and to properly act on the system itself. A low-level tier, instead, contains objects (not necessarily agents) Proceedings 30th European Conference on Modelling and Simulation ©ECMS Thorsten Claus, Frank Herrmann, Michael Manitz, Oliver Rose (Editors) ISBN: 978-0-9932440-2-5 / ISBN: 978-0-9932440-3-2 (CD)