Complex Adaptive Systems and Interactive Granular Computing

Extended Abstract. We discuss an approach to modeling of computations performed by Complex Adaptive Systems (CAS) based on Interactive Granular Computing (IGrC). Complex adaptive systems (CAS) are made up of multiple interacting elements and have the capacity to change and learn from experience. The key problems of complex systems are difficulties with their formal modeling and simulation. Some approaches to modeling CAS are based on agent-based models and/or complex network-based models (see, e.g., [17,36]). Decision support in solving problems related to CAS [6,33,36] requires relevant computation models for the agents as well as methods for incorporating reasoning over computations performed by agents. Agents are performing computations on complex objects (e.g., behavioral patterns, classifiers, clusters, structural objects, sets of rules, aggregation operations, approximate reasoning schemes). In Granular Computing (GrC), all such constructed and/or induced objects are called granules. To model, crucial for CAS, interactive computations [2] performed by agents, we extend the existing GrC approach to Interactive Granular Computing (IGrC) by introducing complex granules (c-granules or granules, for short). Interactive granular computations (computations in IGrC, for short) were proposed as computational models for complex systems. Reasoning about the properties of computations in IGrC are based on adaptive judgment. Computations in this model are performed on c-granules thanks to which it is possible to register, analyze, and synthesize the properties of interactions between physical objects perceived by agents. We assume that the states of certain physical objects, occurring within a specific domain of activity of a given c-granule, are directly measurable. However, the states of other objects are perceived (approximated) indirectly using measurable states, obtained by interactions of physical objects from a particular domain of c-granule’s activity. Each measurable state of a c-granule (at a given moment of the agent’s time) corresponds to a concept. This concept is understood as a set of situations (configurations of physical objects), perceived within this very cgranule, and, thanks to interactions, leads to a specific state. In the proposed app-