This project aims to calibrate both the kinematic parameters and the sensor mounting position of a tricycle robot using real experimental data. A deep explanation can be found here.
The dataset consists of 2434 samples. Each sample includes:
- Time stamps
- Encoder readings for steering and traction wheels
- The robot’s global pose
- The sensor’s global pose
These poses are visualized below:
The tricycle is modeled as a FWD bicycle system.
The calibration focuses on:
- The kinematic parameters
- Sensor pose relative to the robot
The approach uses an iterative least squares algorithm to find the best-fit parameters by minimizing the discrepancy between the measured and predicted sensor movement.
- The error (difference between predicted and measured movement) quickly decreases as the algorithm progresses.
- The number of identified outliers increases during optimization.
- After calibration, the sensor’s measured trajectory (blue) and the calibrated prediction (green) become much more aligned, indicating successful parameter estimation.
Note:
Perfect overlap is not expected due to real-world noise and numerical approximations, but the improvement is clear after calibration.