Odometry

Odometry refers to the estimation of a mobile robot's position and orientation over time by integrating measurements that reflect the robot's own motion. It provides a sequence of relative pose changes, which can be chained to produce a trajectory. Odometry is a foundational component of navigation systems and is often combined with mapping techniques to maintain a consistent global frame.

Wheel odometry uses wheel encoders or similar sensors attached to the drive system to infer translation and rotation from wheel rotations. By applying a kinematic model and wheel radius, the system updates the pose with each encoder reading. Wheel odometry is fast and inexpensive but is sensitive to wheel slip, uneven terrain, and calibration errors, leading to drift over time.

Visual odometry determines motion by analyzing changes in camera images. It can be monocular, stereo, or RGB-D. Feature-based methods track visual features to estimate relative motion, while direct methods use pixel intensities. Visual odometry can recover scale in stereo or RGB-D setups, but monocular systems require additional information. It is robust to wheel slip but sensitive to lighting, texture, and motion blur.

Inertial odometry uses data from inertial measurement units to estimate high-frequency motion by integrating acceleration and angular velocity, but bias and noise cause rapid drift. Sensor fusion techniques, such as extended or unscented Kalman filters, combine inertial data with wheel, visual, or lidar odometry to reduce drift.

In practice, odometry is often fused with simultaneous localization and mapping (SLAM) to provide more accurate and drift-free estimates. Common challenges include accumulating drift, calibration of sensors, scale ambiguity in monocular systems, and varying environmental conditions. Odometry remains essential for real-time navigation, dead reckoning, and initialization for broader mapping systems.