Inertial Navigation Meets Deep Learning: A Survey of Current Trends and Future Directions

Inertial sensing is employed in a wide range of applications and platforms, from everyday devices such as smartphones to complex systems like autonomous vehicles. In recent years, the development of machine learning and deep learning techniques has significantly advanced the field of inertial sensing and sensor fusion, driven by the availability of efficient computing hardware and publicly accessible sensor data. These data-driven approaches primarily aim to enhance model-based inertial sensing algorithms. To foster further research on integrating deep learning with inertial navigation and sensor fusion, and to leverage their potential, this paper presents an in-depth review of deep learning methods in the context of inertial sensing and sensor fusion. We explore learning techniques for calibration and denoising, as well as strategies for improving pure inertial navigation and sensor fusion by learning some of the fusion filter parameters. The reviewed approaches are categorized based on the operational environments of the vehicles—land, air, and sea. Additionally, we examine emerging trends and future directions in deep learning-based navigation, providing statistical insights into commonly used approaches. • Comprehensive review of deep learning methods for inertial sensing and navigation. • Examination of deep learning techniques for calibrating and denoising sensor data. • Analysis of the advantages and disadvantages of deep learning in inertial navigation. • Discussion on future directions for deep learning in inertial sensing and sensor fusion.