Adaptivity in Agent-Based Systems via Interplay between Action Selection and Norm Selection

Beyond the ability for adaptation to an environment, Self-Adaptive Software (SAS) embodies the capacity and the initiative to adapt its own behavior. The adaptation is due to a need to maintain a desired or a reference relationship between a set of input and output signals. We may loosely divide SAS into adaptor and adapted components. The adapted component has an ongoing relationship with the environment. The adaptor detects and evaluates the need for change in the operation of the adapted component. In anthropomorphic terms, this detection and evaluation involves the cognitive processes of introspection and assimilation. However, an artifact may only need a supervisory control module. In a hierarchical system, the idea of adapted and adaptor can be extended to several levels with the higher level adapting the function of lower level. Hierarchical architectural systems are well studied. However, self-adaptive software is more than hierarchical control or application of adaptive techniques such as neural nets or genetic approaches.