In a systematic literature review and bibliometric analysis evaluating learning-based sensor fusion for autonomous mobile robot state estimation, researchers found that digital twin technology has not yet been applied to robot state estimation despite its emerging role.
The review, published in the International Journal of Intelligent Robotics and Applications, formulated ten research questions and retrieved records from Scopus and Web of Science in a single database search conducted on October 24, 2025.
The bibliometric analysis utilized VOSviewer and Bibliometrix to map overall publication trends, tracking the most productive countries, publication sources, and active authors in the domain.
Applying explicit eligibility and quality-assessment criteria, the systematic review analyzed 64 primary studies published between 2019 and 2025 along with 12 digital twin implementations. Cameras and LiDAR were identified as the most frequently fused sensor modalities, while deep learning constituted the most common category, spanning 36 of the 64 primary studies.
The study compared learning-based methods against conventional sensor fusion approaches such as Kalman and particle filters, which can degrade in complex environments, under unmodeled noise, and during highly dynamic motion. Principal method families were analyzed by principle, advantages, limitations, and reported performance, alongside representative studies categorized by fusion level.
Across the 12 analyzed digital twin implementations, only three featured bidirectional data exchange, and none addressed state estimation. Because the integration of digital twins with learning-based sensor fusion remains at an early stage, the authors proposed a research agenda to direct future developments in autonomous mobile robotics.
The research highlights that while traditional techniques like Kalman and particle filters remain foundational, their performance often suffers in challenging operational settings. Specifically, these conventional methods face reliability issues when autonomous mobile robots encounter complex environments, unmodeled noise, or perform highly dynamic maneuvers. Learning-based sensor fusion approaches offer a robust alternative by deriving fusion models directly from gathered data, potentially overcoming the rigid, predefined modeling requirements that limit standard filter performance under unpredictable real-world operating conditions.
To evaluate these technological advancements, the authors employed a bibliometric analysis and a systematic literature review covering scholarly studies published between 2019 and 2025. The bibliometric portion, executed through VOSviewer and Bibliometrix, provided a quantitative map of the current landscape, successfully identifying the most productive countries, primary publication sources, and active researchers contributing to this expanding technical field. By establishing these clear trends, the researchers highlighted the rapid growth and widespread geographic distribution of current scholarly interest in sensor fusion technologies for modern robotics applications.
The systematic review process involved rigorous eligibility and quality-assessment criteria to filter primary studies. By analyzing the 64 primary studies, the researchers compared distinct method families based on their core operating principles, advantages, and limitations. This structured comparison allows for a significantly better understanding of how various fusion levels translate into reported performance metrics. Furthermore, the review categorized studies by fusion level, providing a comprehensive evaluation of how learning-based approaches address the limitations of conventional models. While the inclusion of digital twin technology is still limited and remains in its early stages of integration with robot state estimation, the analysis clearly outlines the current technical barriers and highlights the urgent necessity for further research to bridge existing gaps between current sensor fusion techniques and the prospective use of bidirectional virtual counterparts.
Subject of Research: Learning-based sensor fusion and digital twins for autonomous mobile robot state estimation
Article Title: Learning-based sensor fusion for state estimation in autonomous mobile robots and the emerging role of digital twins: a bibliometric analysis and systematic literature review
Article References: Mirdanies, M., Rashid, R. A., Yazid, E., & Saputra, R. P. (2026). Learning-based sensor fusion for state estimation in autonomous mobile robots and the emerging role of digital twins: a bibliometric analysis and systematic literature review. International Journal of Intelligent Robotics and Applications. https://doi.org/10.1007/s41315-026-00607-x
Image Credits: AI Generated
DOI: 10.1007/s41315-026-00607-x
Keywords: sensor fusion, state estimation, autonomous mobile robots, digital twins, deep learning, systematic literature review, bibliometric analysis, cameras, LiDAR, Kalman filters, particle filters, bidirectional digital twins, Scopus, Web of Science, VOSviewer, Bibliometrix, fusion levels
News Source: Denise Maddox. (November 18, 2026). No Reviewed Digital Twins Address Robot State Estimation, Literature Analysis Finds. Scienmag.



