State Estimation Methods for Autonomous UAVs: A Systematic Review of Kalman-based Particle and Learning Approaches

Abstract

State estimation is a cornerstone of autonomous UAV operation, enabling accurate navigation, control, and mission execution. This systematic literature review (SLR) surveys over 180 peer-reviewed studies on state estimation methods for UAVs, focusing on Kalman filter variants, particle filters, and emerging learning-based approaches. Following PRISMA guidelines, the review categorizes methods by their underlying mathematical models, sensor modalities, and application contexts. Kalman-based filters, including Extended and Unscented Kalman Filters, dominate the literature due to their computational efficiency and well-understood theoretical properties. Particle filters offer advantages in handling non-Gaussian noise and multi-modal distributions but at higher computational costs. Recently, learning-based methods, such as deep neural networks and recurrent architectures, have shown promise in modeling complex sensor dynamics and compensating for unmodeled disturbances. The review critically evaluates each approach’s robustness, scalability, and real-time feasibility, highlighting trade-offs between accuracy and computational demand. Validation practices, including simulation benchmarks and real-world flight tests, are analyzed to assess method maturity. The paper identifies gaps in handling sensor failures, dynamic environments, and long-term drift, suggesting future research directions toward hybrid fusion frameworks that combine model-based and data-driven techniques. This SLR serves as a comprehensive reference for researchers and practitioners seeking to develop resilient UAV state estimation systems.