A Self-Configurable IoT Agent System based on Environmental Variability

This thesis develops a self-configurable system to design agents for Internet of Things (IoT) applications. The proposed approach goes beyond existing methods by supporting handling variability in IoT agents according to environmental changes. As part of the research, we have designed a software framework, prototyped several IoT applications, and conducted simulation and machine learning experiments. We find that (1) IoT agents vary according to the physical, software behavior and analysis architecture; (2) the configuration of the set of agents can be adjusted and reconfigured through feedback evaluative machine learning; and (3) reconfiguring a set of agents dynamically in accordance with environmental variants leads to better performance.