Fuzzy-BDI agents for decision making under uncertainty in smart cyber-physical systems

As embedded system-originated paradigms grow in complexity and lack of autonomy, traditional control architectures struggle to manage the uncertainty and dynamism of real-world environ- ments. Intelligent agent architectures, particularly those grounded in the Belief–Desire–Intention (BDI) model, offer promising cognitive capabilities for Cyber–Physical System (CPS). However, their integration into resource-constrained and (soft) real-time embedded platforms remains chal- lenging. To address this, eventually, we propose a Model-driven Engineering (MDE) approach for deploying fuzzy-BDI agents into CPS. This approach combines fuzzy logic with BDI reason- ing to manage imprecision and context shifts, while also enabling platform-independent system design through model-based abstraction. Despite the growing relevance of such hybrid archi- tectures, existing literature lacks comprehensive methods that link cognitive agent behaviours to deployable embedded code in complex systems settings. This thesis fills this gap by introducing an integrated approach that bridges high-level agent modelling and low-level system implemen- tation. The results demonstrate that fuzzy-BDI agents, supported by model-based analysis and engineering, can enhance adaptability, reduce development complexity, and improve robustness in uncertain environments. The central challenge addressed in this thesis lies in bridging the gap between traditional em- bedded control and intelligent decision-making within CPS. While intelligent agent architec- tures, particularly those grounded in the BDI model, provide a promising cognitive framework for autonomous reasoning, their practical integration with enhanced capabilities into resource- constrained embedded platforms remains an unresolved problem. The BDI model enables agents to reason about their environment, goals, and actions proactively, maintaining beliefs about the world, forming desires representing objectives, and committing to intentions that guide action. However, classical BDI systems are based on crisp logic, which severely limits their applicability under uncertainty and continuous change. Without mechanisms to interpret ambiguous sensory data or adapt to fluctuating contexts, BDI agents fall short of the robustness required in CPS deployments. To address these shortcomings, this thesis introduces an integrated fuzzy-BDI and MDE frame- work for the development and deployment of intelligent CPS. The core idea is to combine fuzzy logic, which allows reasoning with degrees of truth, with the structured deliberation of BDI agents, thereby enabling systems to handle uncertainty natively at every stage from perception and planning to action execution. Complementing this reasoning innovation, the research em- ploys MDE to elevate the abstraction level of system development. The proposed framework bridges high-level agent design models with low-level embedded implementations, achieving ex- plainability, reusability, and platform-independent deployment. This dual integration of fuzzy and BDI reasoning with model-driven design forms the foundation of a new engineering methodology for building adaptive and smart CPS.