kalman_variants implements several Kalman filter nonlinear variants with state estimation applications for robot and autonomous systems in mind. The package includes detailed line-by-line annotated code for every implementation of the kalman filter variant. You will also have free access to an implementation guide at Kalman_filter_variants/doc/Kalman_Implementation_Guide_For_Dummies.pdf to go along with the code to accelerate your learning.

At the moment, this package implements

• Extended Kalman Filter (EKF) w/ landmark-localization tutorial
• Error State Extended Kalman Filter (ESEKF)
• Unscented Kalman Filter (UKF)

Moreover, it implements the following vehicle motion model in both discrete and continuous formulations

• Kinematic Bicycle Model
• Kinematic Differential Drive Model
• Kinematic Ackermann Model
• Dynamic Bicycle Model

• Maintainer Status: Active
• Maintainer(s): Jingxue Jiang
• Author(s): Jingxue Jiang

Download the source via git clone

cd <your directory>
git clone https://github.com/jiangjingxue/kalman_filter_variants.git
python setup.py install

First, import the filters and helper functions.

import numpy as np
from kalman_variants.ekf import EKFLocalization
from kalman_variants.common import AddictdeNoise

Create the filter

my_ekf = EKFLocalization()

if you encounter following issue when importing the package

ModuleNotFoundError: No module named 'kalman_variants'

Set PYTHONPATH environment variable in root project directory (Unix)

export PYTHONPATH=.

For Windows, use

set PYTHONPATH=.