Inertial Navigation and Sensor Fusion Techniques

Inertial navigation relies on motion sensors—accelerometers and gyroscopes—to compute position, orientation and velocity without external references. By integrating sensor outputs over time, inertial navigation systems (INS) track relative movement from a known initial state. High‐performance INS were historically based on stable‐platform designs, but modern strap‐down configurations, in which sensors are rigidly mounted to the moving body, have become prevalent due to advances in micro‐electromechanical systems (MEMS). Sensor fusion techniques address inherent drift by combining inertial measurements with other modalities—satellite navigation, visual odometry, magnetometers or lidar—to constrain accumulated errors. Fusion algorithms range from classical Kalman filters that assume Gaussian noise to non‐linear and adaptive variants capable of handling complex dynamics and environmental changes. Recent trends include machine‐learning frameworks that learn sensor synchronisation and calibration implicitly, and hybrid architectures that allow high‐rate proprioceptive estimation alongside lower‐rate exteroceptive updates. These developments have broadened applications from aerospace and maritime navigation to autonomous vehicles, wearable devices and legged robotics, underlining the global significance of robust, self‐contained positioning solutions.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Self-supervised learning has been applied to monocular visual–inertial odometry, enabling end-to-end estimation of 6‐degree‐of‐freedom motion and scene depth without labelled data. Such frameworks mitigate calibration requirements and improve resilience to sensor misalignment. In legged‐robotics, modular state estimators integrate high‐rate inertial measurements and odometry via extended Kalman filters, while intermittently fusing visual or lidar corrections in a loosely coupled manner to constrain drift over prolonged field missions. Reinforcement‐learning based adaptive filters have also emerged to tune process noise covariances dynamically, yielding robust GNSS/INS localisation under prolonged satellite outages and varying environmental conditions. These diverse approaches demonstrate the trend towards adaptive, data-driven fusion strategies that maintain high‐frequency proprioceptive loops alongside occasional external updates, thus balancing accuracy, robustness and computational efficiency across a spectrum of autonomous systems.

Inertial Navigation and Sensor Fusion Techniques publication trend

The graph below shows the total number of articles in inertial navigation and sensor fusion techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Accelerometer: Sensor measuring linear acceleration along one or multiple axes.

Gyroscope: Sensor measuring angular velocity to infer rotational motion.

Inertial Measurement Unit (IMU): Integrated device combining accelerometers and gyroscopes for motion sensing.

Strap‐down INS: Navigation system in which IMUs are fixed directly to the host structure, requiring computational alignment to inertial frames.

Kalman filter: Algorithm that fuses noisy measurements by estimating state variables under Gaussian noise assumptions.

Extended Kalman filter (EKF): Non‐linear variant of the Kalman filter that linearises around predicted states.

Visual–Inertial Odometry: Technique that combines camera imagery and IMU data to estimate trajectory and orientation.

Reinforcement learning: Machine‐learning approach where an agent learns to optimise decisions via trial‐and‐error and reward feedback.

References

1. Inertial sensors technologies for navigation applications: state of the art and future trends. Satellite Navigation (2020).
2. SelfVIO: Self-supervised deep monocular Visual–Inertial Odometry and depth estimation. Neural Networks (2022).
3. Pronto: A Multi-Sensor State Estimator for Legged Robots in Real-World Scenarios. Frontiers in Robotics and AI (2020).
4. RL-AKF: An Adaptive Kalman Filter Navigation Algorithm Based on Reinforcement Learning for Ground Vehicles. Remote Sensing (2020).