Skip to content

Latest commit

 

History

History
58 lines (43 loc) · 1.87 KB

README.md

File metadata and controls

58 lines (43 loc) · 1.87 KB
Raw

Moment-based Kalman Filter: Nonlinear Kalman Filtering with Exact Moment Propagation

Moment-based Kalman Filter(MKF) is a nonlinear Kalman filter that uses exact moment propagation method to estimate state from noisy measurements.

Paper

This paper has been accepted by the IEEE Conference on Robotics and Automation (ICRA), 2023.

Paper Link

Language and test environment

The main code is written in C++ and visualization code is mainly used Python. However, we also use matplotlib.cpp for the simple visualization in C++. Note that we only test our code on Ubuntu20.04.

Dataset for the experiment

We use UTIAS Multi-Robot Cooperative Localization and Mapping Dataset to evaluate the MKF.

Requirements

CMake >= 3.22

Getting Started

git clone https://github.com/purewater0901/MKF.git
cd MKF
mkdir build && cd build
cmake ..
make -j8
  • Run MKF with Gaussian measurements
./main
  • Run MKF with non-Gaussian measurements
    Before running the code for non-Gaussian environment, you need to preprocess the dataset. After that you can run the main code.
./data_preprocessor
./main_non_gaussian

Result with UTIAS dataset

Robot1

Robot2

References

[1] "Moment-Based Exact Uncertainty Propagation Through Nonlinear Stochastic Autonomous Systems", Ashkan Jasour, Allen Wang, Brian C. Williams https://arxiv.org/abs/2101.12490

[2] "The UTIAS multi-robot cooperative localization and mapping dataset", Leung, Keith and Halpern, Yoni and Barfoot, Timothy and Liu, Hugh, https://journals.sagepub.com/doi/10.1177/0278364911398404