Comprehensive Overview and Comparative Assessment of Model-Based Methods for Vehicle State Estimation

Accurate vehicle state estimation is essential for advanced driver-assistance systems (ADAS) and autonomous driving applications, enabling robust control performance under diverse operating conditions. In this context, it becomes even more critical to design generalizable tools capable of adapting to any vehicle, operating condition, or environment: for this reason, model-based approaches become a true necessity, especially when relying on the outputs of such algorithms in closed-loop control and safety applications, making these methodologies more suitable than purely data-driven approaches. However, in current literature, there exists a variety of model-based estimation strategies with different levels of complexity, accuracy, and computational requirements. This paper reviews approximately 200 studies and presents a comprehensive review of model-based vehicle state estimation techniques, with a focus on their formulation, underlying physical assumptions, and associated trade-offs. Estimation strategies are categorized into pure kinematic, vehicle- and tire-model-based, hybrid, and more complex multibody-model-based approaches, thereby facilitating informed selection of the most suitable method to support design choices based on sensor availability and application requirements. Within each category, a more detailed analysis examines estimation strategies with variable modeling assumptions, observer architectures, and sensor configurations, emphasizing how design choices affect accuracy, robustness to parameter variations, computational cost, and sensitivity to measurement inaccuracies. In contrast to previous surveys, this work provides a structured comparative assessment of all methodologies using specifically defined evaluation criteria, enabling a direct comparison across different estimation approaches and highlighting their respective advantages and limitations. The review concludes by outlining current limitations in the state of the art and identifying promising research directions, including the integration of currently unmodeled physical effects into estimation algorithms, the necessary steps toward simplified calibration and self-calibration procedures even in detailed modeling approaches, and the exploitation of emerging sensor technologies.