Kalman filtering is an algorithm that takes a series of measurements over time, containing statistical noise, and produce an estimate that tends to be more accurate than a single measurement. This is very important, because the the sensors give us always noisy information or the environment could makes data collection difficult , and the predictions made using kalman filtering help or make a better estimate. There are many applications of the kalman filtering, and in this case, it was used to track a spaceship in a simulated trip from the Earth to the Moon.
The trajectory from the Earth to the Moon is a simply "toy" trajectory generated synthetically. Has been generated all the coordinates, velocty and acceleration for a moving object in 3 dimentional space. Then has been added noise with a Gaussian distribuition, in order to simulate a more realistic sensor that provides measurements. This is the trajectory plotted in 3D and in 2D (only x and z axis):
| Kalman Prediction 2D | Kalman Prediction 3D |
|---|---|
 |
 |
The Kalman algorithm is the one implemented in the FilterPy module, and all the implementation choices are described in the report
The following animation shows how the Kalman filtering algorithm works: Takes noisy masurements and then makes an estimate that is better than the measurements.
