• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    迪恩网络公众号

mbrossar/ai-imu-dr: AI-IMU Dead-Reckoning

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称(OpenSource Name):

mbrossar/ai-imu-dr

开源软件地址(OpenSource Url):

https://github.com/mbrossar/ai-imu-dr

开源编程语言(OpenSource Language):

Python 100.0%

开源软件介绍(OpenSource Introduction):

AI-IMU Dead-Reckoning [IEEE paper, ArXiv paper]

1.10% translational error on the KITTI odometry sequences with only an Inertial Measurement Unit.

Results on sequence 08

Overview

In the context of intelligent vehicles, robust and accurate dead reckoning based on the Inertial Measurement Unit (IMU) may prove useful to correlate feeds from imaging sensors, to safely navigate through obstructions, or for safe emergency stop in the extreme case of other sensors failure.

This repo contains the code of our novel accurate method for dead reckoning of wheeled vehicles based only on an IMU. The key components of the method are the Kalman filter and the use of deep neural networks to dynamically adapt the noise parameters of the filter. Our dead reckoning inertial method based only on the IMU accurately estimates 3D position, velocity, orientation of the vehicle and self-calibrates the IMU biases. We achieve on the KITTI odometry dataset on average a 1.10% translational error and the algorithm competes with top-ranked methods which, by contrast, use LiDAR or stereo vision.

Structure of the approach

The above figure illustrates the approach which consists of two main blocks summarized as follows:

  1. the filter integrates the inertial measurements with exploits zero lateral and vertical velocity as measurements with covariance matrix to refine its estimates, see the figure below;
  2. the noise parameter adapter determines in real-time the most suitable covariance noise matrix. This deep learning based adapter converts directly raw IMU signals into covariance matrices without requiring knowledge of any state estimate nor any other quantity.

Structure of the filter

Code

Our implementation is done in Python. We use Pytorch for the adapter block of the system. The code was tested under Python 3.5.

Installation & Prerequies

  1. Install pytorch. We perform all training and testing on its development branch.

  2. Install the following required packages, matplotlib, numpy, termcolor, scipy, navpy, e.g. with the pip3 command

pip3 install matplotlib numpy termcolor scipy navpy
  1. Clone this repo
git clone https://github.com/mbrossar/ai-imu-dr.git

Testing

  1. Download reformated pickle format of the 00-11 KITTI IMU raw data at this url, extract and copy then in the data folder.
wget "https://www.dropbox.com/s/ey41xsvfqca30vv/data.zip"
mkdir ai-imu-dr/results
unzip data.zip -d ai-imu-dr
rm data.zip

These file can alternatively be generated after download the KITTI raw data and setting read_data = 1 in the main.py file.

  1. Download training parameters at this url, extract and copy in the temp folder.
wget "https://www.dropbox.com/s/77kq4s7ziyvsrmi/temp.zip"
unzip temp.zip -d ai-imu-dr/temp
rm temp.zip
  1. Test the filters !
cd ai-imu-dr/src
python3 main_kitti.py

This first launches the filters for the all sequences. Then, results are plotted. Note that the parameters are trained on sequences 00, 01, 04-11, so only sequence 02 is a test sequence.

Training

You can train for testing another sequence (we do not find difference in the results) or for our own sequence by modifying the dataset class.

Paper

The paper M. Brossard, A. Barrau and S. Bonnabel, "AI-IMU Dead-Reckoning," in IEEE Transactions on Intelligent Vehicles, 2020, relative to this repo is available at this url.

Citation

If you use this code in your research, please cite:

@article{brossard2019aiimu,
  author = {Martin Brossard and Axel Barrau and Silv\`ere Bonnabel},
  journal={IEEE Transactions on Intelligent Vehicles}, 
  title = {{AI-IMU Dead-Reckoning}},
  year = {2020}
}

Authors

Martin Brossard*, Axel Barrau° and Silvère Bonnabel*

*MINES ParisTech, PSL Research University, Centre for Robotics, 60 Boulevard Saint-Michel, 75006 Paris, France

°Safran Tech, Groupe Safran, Rue des Jeunes Bois-Châteaufort, 78772, Magny Les Hameaux Cedex, France




鲜花

握手

雷人

路过

鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
热门推荐
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服(服务时间 9:00~18:00)

在线QQ客服
地址:深圳市南山区西丽大学城创智工业园
电邮:jeky_zhao#qq.com
移动电话:139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap