A Theoretical and Methodological Review of the Data Fusion Process: Architectures, Algorithms, and Challenges

This study provides a comprehensive theoretical and methodological analysis of the data fusion process, systematically addressing its application levels, algorithms, architectural models, and associated challenges. Data fusion plays a pivotal role across various disciplines—including healthcare, agriculture, environmental monitoring, autonomous systems, robotics, and industrial domains—by generating more reliable information, reducing uncertainty, and strengthening decision-support mechanisms. In the literature, data fusion approaches are classified into rule-based, probabilistic, artificial intelligence-driven, and optimization-oriented methods, each offering distinct advantages and limitations depending on the scenario and data type. Furthermore, centralized, distributed, and hybrid architectures are evaluated in terms of scalability, fault tolerance, and real-time performance. This study also highlights critical challenges in data fusion, such as data heterogeneity, alignment issues, hardware constraints, and privacy and security concerns, while exploring strategies to address these barriers. Ultimately, data fusion is positioned not merely as a technical integration process but as a strategic methodological framework for building sustainable and trustworthy decision-support systems.