Robust Cooperative localization simulation

A simulation code for the paper "Robust Cooperative Localization with Failed Communication and Biased Measurements",
published in IEEE Robotics and Automation Letters with a supplemental video that is also in bilibili
View Demo · Report Bug OR Request Feature

Table of Contents

1. Introduction
   • Getting Started
     • Usage
     • Structure
     • Roadmap
     • Errata
     • License
     • Contact
     • Acknowledgments

Introduction

Cooperative Localization (CL) plays a crucial role in achieving precise localization without relying on localization sensors. However, the performance of CL can be significantly affected by failed communication and biased measurements. This code shows a robust decentralized CL method that addresses these challenges effectively.

Please cite this work when referencing:

@ARTICLE{RobustCL_RAL,
  author={He, Ronghai and Shan, Yunxiao and Huang, Kai},
  journal={IEEE Robotics and Automation Letters}, 
  title={Robust Cooperative Localization With Failed Communication and Biased Measurements}, 
  year={2024},
  volume={9},
  number={3},
  pages={2997-3004}
}

About this paper: there are a few mistakes shown in the Errata part.

Getting Started

1. Since this is a Python program, you will need Python3. Please Make sure you have it installed.
2. Simply clone the repo

   git clone https://github.com/RonghaiHe/RobustCL.git

3. Install the Python libraries that this program depends on through requirements.txt:

   pip install -r requirements.txt

Usage

• If runinng the program with specific failed communication probability(e.g. 0.5):

  python3 main_fix.py --comm_fail_prob 0.5

• If runinng the program with multiple failed communication probability simultaneously:

  bash shell_main_fix.sh

• If runinng the mission(e.g. 1) with multiple failed communication probability simultaneously:

  bash shell_main.sh 1

  Details of missions are:

  0 Draw the sample trajectory, for Response;

  1 ARMSE over time, for Fig. 1 & Fig. 2 in the paper;

  2 weight of M-estimation over time, for Fig. 3 in the paper;

  3 ARMSE over tau, for Fig. 4 in the paper;

  4 animation about ARMSE over time, for videos about the paper;

  5 animation about ARMSE over tau, for videos about the paper

• If runinng the mission(e.g. 1) with multiple failed communication probability simultaneously in the original terminal, try using tmuxp after installing tmux:

  MISSION=1 tmuxp load robust_cl.yaml

Structure

Repository contains:

• algorithms, a directory containing all algorithms used et al, off the record temporarilyup to date.
  • BDA.py: "BDA" algorithm in the paper;
  • CI.py: "CI" algorithm in the paper;
  • CU.py: "BDA-CU", "CI+CU" algorithms in the paper;
  • DCL_GS.py: Our proposed method in the paper;
  • DMV.py: "DMV" algorithm in the paper;
  • DR.py: "DR" method in the paper;
  • func_detect_fault.py: multiple functions about detecting the fault measurement;
• others, a directory containing other calculation in the paper
  • KL_calc.py: calulate the numerical KL-divergence;
  • video_output.py: multiple figures -> video;
• draw.py: draw the final figures;
• main_fix.py: main function. Set the fix running parameters here;
• parameters.py: sets the system parameters;
• requirements.txt, a file containing all Python libraries needed;
• shell_main.fix.sh: a shell file to run the program simultaneously;
• utils.py: execute the whole simulation

Roadmap

• Whole codes uploaded
• Formatted with pep8
• Complete Python project structure

See Changelog.md for the change at each version

Errata

The conclusion of the paper isn't wrong, whereas there are a few mistakes in this paper. Sorry about that and declare here:

1. In formulas (17) and (20), the dimension of $\boldsymbol{I}$ should be $3N\times 3N$ rather than $N\times N$ (By Chang Liu)
2. Formula (26) is wrong, $\boldsymbol{S}_ {t+1} \to \boldsymbol{M}_{t+1}$ (By Chang Liu)
3. (Confusion Part)In the formula (23), the variable $\hat{\boldsymbol{X}}_ {t+1 \mid t}^i$ is shown for explanation. During the iteration, it is replaced by the result after $n$-th iterations. Specifically, we use the latest iteration result, X_all, to calculate the residual item in this code (By Chang Liu)
4. The serial numbers of references used for comparison in Fig. 1, Fig. 2, Fig. 4 and TABLE II haven't been updated. The right serial numbers can be seen in the last paragraph of section V-A: DMV [13], BDA [12], BDA-CU [4] and CI+CU [17] (By a professor).

Still updating ...

License

Distributed under the MIT License. See LICENSE for more information.

Contact

RonghaiHe - E-mail

Project Link: RobustCL

Acknowledgments

• All authors of this paper;
• Readers pointing out errors or improvements about the paper: professors, Chang liu;
• Inspired by this repository;
• A recommended template. This README.md is modified from that.