Wheel Odometry

Wheel odometry uses physical sensors on the wheels to measure rotation and calculate distance. Visual odometry uses cameras to track key points in the environment to estimate movement. While visual odometry is better at correcting long-term drift, wheel odometry is faster, computationally cheaper, and works in the dark or featureless environments.

Drift occurs because errors accumulate. Tiny inaccuracies caused by wheel slippage, uneven floors, slightly unequal tire pressures, or imperfect wheel alignment are integrated over time. Without an external reference (like LiDAR or GPS) to "reset" the position, a 1% error can result in being meters away from the target after a long run.

Quadrature incremental optical encoders are the industry standard for AGVs. They provide high resolution (thousands of ticks per revolution) and direction sensing. For absolute positioning upon startup without homing, magnetic absolute encoders are sometimes used, though they are more expensive.

The wheel diameter is a constant in the kinematic formula. If the physical wheel wears down (decreasing diameter) or compresses under load, the robot travels less distance per rotation than the software expects. Regular calibration of the wheel radius parameter is essential for maintaining accuracy.

Yes, but the math is more complex. Mecanum wheels allow omnidirectional movement, so the kinematic model must account for the rollers' vector forces. Furthermore, Mecanum wheels generally experience more slippage than standard tires, making odometry data noisier and less reliable over long distances.

For modern mobile robots, odometry is typically calculated at 50Hz to 100Hz (50 to 100 times per second). High-speed robots may require faster rates. This high frequency ensures smooth control loops and allows the robot to react quickly to velocity changes.

Pure wheel odometry cannot detect slippage (e.g., spinning on oil). To handle this, engineers use Sensor Fusion (often via an Extended Kalman Filter). The filter compares wheel data with IMU (accelerometer/gyro) data; if the wheels spin but the IMU detects no acceleration, the system identifies the slip and relies on the IMU temporarily.

Rarely on its own. It is sufficient for very short movements or straight-line blind moves. For full autonomy, it is almost always paired with LiDAR (SLAM), Visual SLAM, or magnetic tape/QR codes to correct the inevitable accumulated error (drift) relative to the global map.

The wheelbase is the distance between the contact points of the left and right wheels. It is critical for calculating rotation. If the software's wheelbase value differs from the physical reality (even by millimeters), every turn the robot makes will introduce a systematic orientation error.

A common method is the UMBmark test. The robot is programmed to drive a square path (4 meters per side). By measuring the offset between the start and end point after the square is completed, you can calculate correction factors for both wheel diameter and wheelbase width to tune the kinematic model.

Significantly. Soft carpets increase resistance and may cause micro-slippage. Polished concrete can be slippery. Uneven transitions or bumps can cause wheels to momentarily lose contact with the ground. Hard, flat, high-friction surfaces like concrete or industrial tile provide the best results.

In ROS (Robot Operating System), odometry is broadcast as a standard message type containing the Pose (position/orientation) and Twist (linear/angular velocity). It is also broadcast as a coordinate transform (TF) from the `odom` frame to the `base_link` frame.