This python package implements algorithms for probabilistic state estimation (e.g., Kalman Filter, Particle Filter). So far it only includes a Kalman Filter implementation module. Tested and working in Linux Ubuntu 18.06.
- numpy
- scipy
To install the package, you may use PIP:
pip install -U stateEstimation
Alternatively, you can install the package locally. First, clone the repository to your computer:
git clone http://github.com/rgmaidana/stateEstimation $HOME/stateEstimation
Go to the cloned directory and run:
pip install -e .
To use the package, simply import it in python:
import stateEstimation
The Kalman Filter module implements the linear Kalman Filter algorithm. To use it, the user must instantiate the module with the state-space matrices (i.e., A and B), as well as the measurement model matrix (H), the model uncertainty matrix (Q), and the sensor covariance matrix (R).
Then the user must call the run() function to perform the prediction and update steps. For example:
from stateEstimation import KF
# Define the state-space matrices, i.e. A, B
# Define the Kalman Filter matrices, i.e. H, Q, R
# Instantiate Kalman Filter
filter = KF(A, B, H, Q, R)
filter.run()
An example of the state estimation of a simulated DC motor can be found in the examples folder. The module may also be used as an Extended Kalman Filter (EKF), an Unscented Kalman Filter (UKF), etc, as the state-space and filter matrices are user-defined (i.e., the state-space and measurement model matrices may be linearized by the user). An example of the filter being used as an EKF to estimate the angle and velocity of a simple pendulum can be found in the examples.
In both examples, normal gaussian noise is used to demonstrate the effects of sensor measurement variance in the state estimation. The error magnitude can be adjusted in the example scripts (e.g., 0 for no error).