Context-sensitive Modeling of Input Source Configuration for Evolving Intelligent Systems Dominik Wachholder Department of Information Systems Communications Engineering Kepler University Linz, Austria Email: dominik.wachholder@jku.at Chris Stary Department of Information Systems Communications Engineering Kepler University Linz, Austria Email: christian.stary@jku.at Abstract—Today’s dynamic in system interaction and user behavior is crucial for quality of operation and performance of evolving intelligent systems. Changes in network topology and in- formation availability need to be made accessible instantaneously not only for evolving intelligent systems but also for people using and configuring these systems in ubiquitous computing environments. To insure interoperability among these systems, we present a representation and runtime approach for modeling input source configuration aligned to both user and automatic system invocation. It is based on input sources that are arranged dynamically in their context using bigraphs. This particular rep- resentation approach enforces rigorous interoperable semantics using a high-level set of dynamic relationships. We demonstrate the effectiveness of bigraph-based modeling for dynamic smart home control. I. I NTRODUCTION Today’s organization of work and life is increasingly re- quiring flexible and dynamic designs, not only with respect to how activities and tasks are organized but also regarding evolving intelligent systems (eIS) that need to be responsive to modifications emerging at runtime [1], [2]. This is especially true when people have to be able to use and configure such systems in ubiquitous computing environments without being cognitively and physically overloaded [3]. Consider an ever- changing and unpredictable network of sensors and devices for smart home control, including temperature sensors, barometers and smart phones. Evolving intelligent systems are required to be able to take into account all information available on the network, irrespective of whether a new input source emerges (e.g., smart phone) or information of a new type is provided (e.g., camera). Such systems need to continuously evolve by means of adapting their input source configuration and reasoning model to improve quality of operation and performance [4], [5]. Evolvability of (intelligent) systems concerns the ability to “alter their structure or function so as to adapt to changing circumstances” [6], [7]. Hereby, such a system should keep some identity, “while having the capability to retain certain characteristics of interests despite changes in its composition, topology, and environmental properties. The latter is termed as robustness. In case changes are dynamically tolerated, a system is termed as resilient. It should tolerate (or even profit from) the onset of unanticipated changes and environmental conditions that might otherwise cause a loss of acceptable service” [8]. Consequently, evolving 1 intelligent systems should be resilient and robust as they develop in non-stationary environments (cf. [5]). In the context of evolving fuzzy systems, an important sub-area of evolving intelligent systems, these characteristics are referred to as stability and plasticity [9]. The focus here is on finding the balance between stable converging solutions (stability) while fostering the ability to learn, that is to integrate new information (plasticity) 2 . S B S A (a) System Integration S B S A (b) System Interoperability Figure 1: System Integration vs. Interoperability For learning in non-stationary environments, evolving in- telligent systems build upon a data-driven approach [1]. The foundation of operation is based on autonomous external input sources providing continuous data streams dynamically arranged around the system. Efficient orchestration of het- erogeneous input sources, however, requires not only their consolidation, but also semantic interoperability to achieve a common goal [11]. Orchestrated systems therefore need to keep operating separately without having functional dependen- cies among each other, but still are required to have the ability to cooperate with each other in order to accomplish a common goal. Such systems are characterized as interoperable as they agree upon a common way of interaction in order to collaborate and exchange information (cf. [12]). As such, they are distinct from integrated systems – see Figure 1 showing two systems S A and S B . In the first case (cf. Figure 1a), both systems are integrated as they share common parts of their functionality 1 In the context of evolving intelligent systems, the term evolvable rather pursues the idea of gradual system development than making use of concepts from evolutionary algorithms, such as mutation and selection. 2 The endeavor of finding the balance between stability and plasticity is also referred to as stability-plasticity dilemma (cf. [10]).