Modern autonomous and intelligent vehicles increasingly require reliable in-cabin monitoring systems capable of interpreting complex physiological and behavioral signals to safeguard driver well-being. This paper proposes a time-aware, ontology-driven framework that unifies multimodal physiological, behavioral, and cognitive indicators into a single semantic model capable of transparent, rule-based driver-state inference. The proposed system combines GCIs for interpretable physiological threshold classification with SWRL rules for contextual and numeric reasoning, enabling transparent inference of fatigue, attention level, and unresponsiveness across sequential time windows. A 4D Description Logic architecture structures each measurement as a temporally indexed state, supporting longitudinal analysis over time-series data. To ensure robustness and scalability, real-world physiological data are augmented with synthetic physiological time series generated via TimeGAN, while the inferred driver states condition a diffusion-based image generation module for visual interpretation. The framework is evaluated through a comprehensive validation stack, including competency questions, structural consistency checks, performance profiling, image quality assessment using a self-supervised ARNIQA model and cross-model consistency using visual extractors with an LLM, demonstrating high logical consistency, accurate state inference, and stable performance across multiple distinct actors with 16 temporal slices. Results confirm that the proposed ontology-based framework provides a transparent, extensible, and computationally tractable foundation for in-cabin driver state assessment, enabling future integration with advanced Human-Machine Interface (HMI) systems and real-time intelligent vehicle applications.