The rapid expansion of the Internet of Things (IoT), amplified by the integration of Artificial Intelligence (AIoT), presents significant architectural complexities for software system design. A critical gap persists in providing architects with accessible, evidence-based guidance to navigate these challenges, often leading to suboptimal designs and project failures. This dissertation addresses this gap by investigating the core research question: "What IoT application domains and characteristics of their software systems architectures influence Quality Requirements (QRs) and how this knowledge can be systematically organized and offered to support the decision-making in IoT software systems projects?"
To answer this question, a Systematic Literature Review (SLR) was conducted, analyzing 37 primary studies to distill actionable architectural knowledge. The primary contribution of this work is twofold: first, the creation of a comprehensive and structured Knowledge Base of IoT architectural solutions; and second, the development of ArchIoTec, a novel decision-support tool. ArchIoTec provides a dual-modality interface, allowing users to explore the knowledge base through both hierarchical browsing and a conversational AI assistant powered by a Retrieval-Augmented Generation (RAG) architecture. This AI grounds its responses exclusively in the curated knowledge base, ensuring domain-specific accuracy.
The tool's effectiveness, efficiency, and utility were validated through an evaluation involving realistic design scenarios tailored for software architects and engineers. The results demonstrate that ArchIoTec successfully provides relevant and actionable guidance. This research culminates in a tangible, knowledge-driven tool that bridges the gap between fragmented academic theory and industry practice, empowering architects to make more informed and effective design decisions for complex IoT systems.
.png)
.png)
