Practical Implementation of KalmanNet for Accurate Data Fusion in Integrated Navigation

Introduction

[Paper Link]

1. Song, J., Mei, W., Xu, Y., Fu, Q. & Bu, L. Practical implementation of KalmanNet for accurate data fusion in integrated navigation. IEEE Signal Process. Lett. 1–5 (2024) doi:10.1109/LSP.2024.3431443.

Abstract

The extended Kalman filter has been widely used in sensor fusion to achieve integrated navigation and localization. Efficiently integrating multiple sensors requires prior knowledge about their errors for setting the filter. The recently emerged KalmanNet managed to use recurrent neural networks to learn prior knowledge from data and carry out state estimation for problems under non-linear dynamics with partial information. In this letter, the KalmanNet is implemented for integrated navigation using data from GPS/Wheels and the Inertial Measurement Unit. Therein, a practical strategy for the training algorithm of truncated backpropagation through time is presented by taking advantage of the first-order Markov property of the system state of the Kalman filter, which improves the training robustness and performance of the existing KalmanNet. Experimental results on the Michigan NCLT dataset show that our fusion KalmanNet significantly outperforms the conventional EKF-based fusion algorithm with an improvement of 20% ∼ 40% in average RMSE.

Flow Chart

2012-11-16 Result on Map

Environment Setup

pip install -r requirements.txt

Data Preprocessing

1. We have provided data in the ./data/NCLT/processed/ that has been preprocessed (consistent with the paper).
2. If you want to re-download the data and process it, use the following steps

# download sensor data and ground truth
python ./data/NCLT/down.py --sen 
python ./data/NCLT/down.py --gt 
ls ./data/NCLT/download/sensor_data/*.tar.gz | xargs -n1 tar xzvf
python ./data/NCLT/preprocess.py

Train and Predict

1. Train.

# KalmanNet
python train.py --cfg ./configs/nclt/fusion/wheel_gpsfusion_origin.py
# Split KalmanNet
python train.py --cfg ./configs/nclt/fusion/wheel_gpsfusion_split.py

2. Inference.
   Reference nclt_predict.ipynb

Plotting on Map

We recommend using QGIS for visualization on maps. If you only want to see our results: Using QGIS to open ./QGIS/20121116.qgz. Else if you want to plot your results:

1. Using the code provided at the bottom of nclt_predict.ipynb, the coordinates are converted and saved into kml (Code is modified from https://github.com/AbhinavA10/mte546-project).
2. Open QGIS and load vector(xxx.kml)

Thanks

We are very grateful to the author of https://github.com/AbhinavA10/mte546-project which has inspired us a lot!

Citation

If you find the paper/code useful, please cite our paper using the following BibTeX entry.

@ARTICLE{10605082,
  author={Song, Jian and Mei, Wei and Xu, Yunfeng and Fu, Qiang and Bu, Lina},
  journal={IEEE Signal Processing Letters}, 
  title={Practical Implementation of KalmanNet for Accurate Data Fusion in Integrated Navigation}, 
  year={2024},
  volume={31},
  number={},
  pages={1890-1894},
  keywords={Training;Sensor fusion;Global Positioning System;Navigation;Vectors;Kalman filters;Wheels;Integrated navigation and localization;Kalman filter;recurrent neural networks;sensor fusion},
  doi={10.1109/LSP.2024.3431443}}