Understanding why SLAM algorithms fail in modern indoor environments

Nwankwo Linus* · Elmar Rueckert

arXiv | Paper | Dataset

LiDAR-based data acquisition	Visual-based data acquisition

Table of Contents

• Introduction
• Setup Instruction
• Run the example files
• Citation

Introduction

This repository contains the resources for the paper "Understanding Why SLAM Algorithms Fail in Modern Indoor Environments". The paper investigates the reasons for the failures of simultaneous localization and mapping (SLAM) algorithms in modern indoor environments and provides insights into improving their performance.

Setup Instruction

• Set up ROS environment:

To run the code and reproduce the experiments conducted in the paper, you need to set up a ROS (Robot Operating System) environment. Follow the official ROS documentation here to install ROS on your system. We used ROS Noetic distribution for all our experiments. Make sure you have the necessary dependencies installed.

• Create a workspace (optional):

I called my workspace o2s workspace (o2s_ws). Yours may be different.

mkdir -p ~/o2s_ws/src
cd ~/o2s_ws/src
cd ..
catkin_make

Set path of the workspace as follows:

source devel/setup.bash
gedit ~/.bashrc

Add source ~/o2s_ws/devel/setup.bash to the very end of the opened bash file so that you don't need to source the bash file each time you open a new terminal window.

• Download datasets:

The datasets are recorded in ROS bag files for each environment. The data include the LiDAR, RGB-D and IMU information from the sensors. To analyze the data and reproduce the experiments, you will need to download the ROS bag files containing the sensor data from each of the indoor environments. You can download the ROS bag files from here. Once downloaded, place the ROS bag files in the workspace (src) directory. Before proceeding to the next step, make sure that you have the relevant SLAM algorithms set up on your system. Lidar-based: Hector-SLAM, Gmapping and Karto-SLAM. Visual-based: RTAB and ORB-SLAM2. To run any of the SLAM algorithms on the downloaded dataset, follow the following steps🥇:

roscore

starts the ROS core node. Check if the ROS bag contains all the necessary ROS topics.

rosbag info lidarHospitalEnv.bag

For lidar-based SLAM using Hector-SLAM for example, execute the following

rosparam set use_sim_time true
rosbag play bagfiles/HospitalEnv/lidarHospitalEnv.bag --clock
roslaunch hector_slam_launch tutorial.launch
rosbag record -O lidarHospitalEnvData.bag -a

For visual-based datasets using RTAB for example, execute the following

rosparam set use_sim_time true
rosbag play /bagfiles/HospitalEnv/visualHospitalEnv.bag  --clock
roslaunch rtabmap_ros rtabmap.launch rtabmap_args:="--delete_db_on_start"
rosbag record -O visualHospitalEnvData.bag -a

• Extract useful data from the recorded ROS bag file:

To extract the relevant topics from the recorded rosbag files, you need to run the rosbagPOSEtoCSV.py or rosbagTFtoCSV.py scripts. This script is included in the repository and provides functionality to extract the tf, odom, pose, etc from the rosbag files. Execute the following command in your terminal:

python3 rosbagPOSEtoCSV.py

Note, before running the above code, do not forget to replace /home/linus/o2s_ws/src/o2s_robot/bagfile/raadDataset/controlledEnv/lidar/bags/gtCtrldEnv.bag with the actual path to the ROS bag file you want to extract topics from. Do the same for the output file path ( /home/linus/o2s_ws/src/o2s_robot/bagfile/raadDataset/controlledEnv/lidar/gtCtrldEnv.csv ). The script will extract the desired topics and save them as CSV files for further analysis. Once all the relevant topics are extracted and saved in CSV format, follow the following examples to replicate our experiments.

Run the example files

• Requirements:

To run the example codes provided in this repository, you need to have Matlab installed on your system. Make sure you have a compatible version of Matlab (at least version 2020a) and the required toolboxes installed. The example codes utilize Matlab functionalities for data analysis and visualization.

• Clone the GitHub repository:

To access the example files and scripts used in the paper, clone this GitHub repository to your local PC. Use the following command:

cd ~/o2s_ws/src
git clone https://github.com/LinusNEP/RAAD-2023.git

This will create a local copy of the repository on your PC, allowing you to run the example files and explore the code.

• Run and evaluate:

The example Matlab codes run in sections, first you have to use the 'Run' command to load the parameters to the Matlab workspace, it will return an interactive dialogue box where you have to select yes to load the data to the workspace. Subsequently, for any of the environments, select the metrics of choice (e.g., show the trajectories, absolute trajectory error (ATE), etc) for which you wish to analyse.

Please refer to the paper for more details on the experiments, methodology, and analysis conducted in this work.

Citation

If you use this work for any academic task, please cite the paper.

@InProceedings{10.1007/978-3-031-32606-6_22,
author="Nwankwo, Linus
and Rueckert, Elmar",
editor="Petri{\v{c}}, Tadej
and Ude, Ale{\v{s}}
and {\v{Z}}lajpah, Leon",
title="Understanding Why SLAM Algorithms Fail in Modern Indoor Environments",
booktitle="Advances in Service and Industrial Robotics",
year="2023",
publisher="Springer Nature Switzerland",
address="Cham",
pages="186--194"
}