SSL_SLAM2: Lightweight 3-D Localization and Mapping for Solid-State LiDAR (mapping and localization separated) ICRA 2021

Overview

SSL_SLAM2

Lightweight 3-D Localization and Mapping for Solid-State LiDAR (Intel Realsense L515 as an example)

This repo is an extension work of SSL_SLAM. Similar to RTABMAP, SSL_SLAM2 separates the mapping module and localization module. Map saving and map optimization is enabled in the mapping unit. Map loading and localization is enabled in the localziation unit.

This code is an implementation of paper "Lightweight 3-D Localization and Mapping for Solid-State LiDAR", published in IEEE Robotics and Automation Letters, 2021 paper

A summary video demo can be found at Video

Modifier: Wang Han, Nanyang Technological University, Singapore

Running speed: 20 Hz on Intel NUC, 30 Hz on PC

1. Solid-State Lidar Sensor Example

1.1 Scene reconstruction example

1.2 Localization with built map

1.3 Comparison

2. Prerequisites

2.1 Ubuntu and ROS

Ubuntu 64-bit 18.04.

ROS Melodic. ROS Installation

2.2. Ceres Solver

Follow Ceres Installation.

2.3. PCL

Follow PCL Installation.

Tested with 1.8.1

2.4. GTSAM

Follow GTSAM Installation.

2.5. Trajectory visualization

For visualization purpose, this package uses hector trajectory sever, you may install the package by

sudo apt-get install ros-melodic-hector-trajectory-server

Alternatively, you may remove the hector trajectory server node if trajectory visualization is not needed

3. Sensor Setup

If you have new Realsense L515 sensor, you may follow the below setup instructions

3.1 L515

3.2 Librealsense

Follow Librealsense Installation

3.3 Realsense_ros

Copy realsense_ros package to your catkin folder

    cd ~/catkin_ws/src
    git clone https://github.com/IntelRealSense/realsense-ros.git
    cd ..
    catkin_make

4. Build SSL_SLAM2

4.1 Clone repository:

    cd ~/catkin_ws/src
    git clone https://github.com/wh200720041/ssl_slam2.git
    cd ..
    catkin_make
    source ~/catkin_ws/devel/setup.bash

4.2 Download test rosbag

You may download our recorded data: MappingTest.bag (3G) and LocalizationTest.bag (6G)if you dont have realsense L515, and by defult the file should be under home/user/Downloads

unzip the file (it may take a while to unzip)

cd ~/Downloads
unzip LocalizationTest.zip
unzip MappingTest.zip

4.3 Map Building

map optimization and building

    roslaunch ssl_slam2 ssl_slam2_mapping.launch

The map optimization is performed based on loop closure, you have to specify the loop clousre manually in order to trigger global optimization. To save map, open a new terminal and

  rosservice call /save_map

Upon calling the serviece, the map will be automatically saved. It is recommended to have a loop closure to reduce the drifts. Once the service is called, loop closure will be checked. For example, in the rosbag provided, the loop closure appears at frame 1060-1120, thus, when you see "total_frame 1070" or "total_frame 1110" you may immediately type

  rosservice call /save_map

Since the current frame is between 1060 and 1120, the loop closure will be triggered automatically and the global map will be optimized and saved

4.4 Localization

Type

    roslaunch ssl_slam2 ssl_slam2_localization.launch

If your map is large, it may takes a while to load

4.5 Parameters Explanation

The map size depends on number of keyframes used. The more keyframes used for map buildin, the larger map will be.

min_map_update_distance: distance threshold to add a keyframe. higher means lower update rate. min_map_update_angle: angle threshold to add a keyframe. higher means lower update rate. min_map_update_frame: time threshold to add a keyframe. higher means lower update rate.

4.6 Relocalization

The relocalization module under tracking loss is still under development. You must specify the robot init pose w.r.t. the map coordinate if the starting position is not the origin of map. You can set this by

    <param name="offset_x" type="double" value="0.0" />
    <param name="offset_y" type="double" value="0.0" />
    <param name="offset_yaw" type="double" value="0.0" />

4.7 Running speed

The realsense is running at 30Hz and some computer may not be able to support such high processing rate. You may reduce the processing rate by skipping frames. You can do thid by setting the

<param name="skip_frames" type="int" value="1" />

1 implies no skip frames, i.e., 30Hz; implies skip 1 frames, i.e., 15Hz. For small map building, you can do it online. however, it is recommended to record a rosbag and build map offline for large mapping since the dense map cannot be generated in real-time.

5 Map Building with multiple loop closure places

5.1 Dataset

You may download a larger dataset LargeMappingTest.bag (10G), and by defult the file should be under home/user/Downloads

unzip the file (it may take a while to unzip)

cd ~/Downloads
unzip LargeMappingTest.zip

5.2 Map Building

Two loop closure places appear at frame 0-1260 and 1270-3630, i.e., frame 0 and frame 1260 are the same place, frame 1270 adn 3630 are the same place. Run

    roslaunch ssl_slam2 ssl_slam2_large_mapping.launch

open a new terminal, when you see "total_frame 1260", immediately type

  rosservice call /save_map

when you see "total_frame 3630", immediately type again

  rosservice call /save_map

6. Citation

If you use this work for your research, you may want to cite the paper below, your citation will be appreciated

@article{wang2021lightweight,
  author={H. {Wang} and C. {Wang} and L. {Xie}},
  journal={IEEE Robotics and Automation Letters}, 
  title={Lightweight 3-D Localization and Mapping for Solid-State LiDAR}, 
  year={2021},
  volume={6},
  number={2},
  pages={1801-1807},
  doi={10.1109/LRA.2021.3060392}}
Owner
Wang Han 王晗
I am currently a Phd Candidate at Nanyang Technological University, Singapore, specialize in computer vision and robotics
Wang Han 王晗
Strongly local p-norm-cut algorithms for semi-supervised learning and local graph clustering

Strongly local p-norm-cut algorithms for semi-supervised learning and local graph clustering

Meng Liu 2 Jul 19, 2022
Tools for robust generative diffeomorphic slice to volume reconstruction

RGDSVR Tools for Robust Generative Diffeomorphic Slice to Volume Reconstructions (RGDSVR) This repository provides tools to implement the methods in t

Lucilio Cordero-Grande 0 Oct 29, 2021
More Photos are All You Need: Semi-Supervised Learning for Fine-Grained Sketch Based Image Retrieval

More Photos are All You Need: Semi-Supervised Learning for Fine-Grained Sketch Based Image Retrieval, CVPR 2021. Ayan Kumar Bhunia, Pinaki nath Chowdh

Ayan Kumar Bhunia 22 Aug 27, 2022
This codebase is the official implementation of Test-Time Classifier Adjustment Module for Model-Agnostic Domain Generalization (NeurIPS2021, Spotlight)

Test-Time Classifier Adjustment Module for Model-Agnostic Domain Generalization This codebase is the official implementation of Test-Time Classifier A

47 Dec 28, 2022
PyTorch implementation of saliency map-aided GAN for Auto-demosaic+denosing

Saiency Map-aided GAN for RAW2RGB Mapping The PyTorch implementations and guideline for Saiency Map-aided GAN for RAW2RGB Mapping. 1 Implementations B

Yuzhi ZHAO 20 Oct 24, 2022
Bayesian optimization in PyTorch

BoTorch is a library for Bayesian Optimization built on PyTorch. BoTorch is currently in beta and under active development! Why BoTorch ? BoTorch Prov

2.5k Dec 31, 2022
Datasets, Transforms and Models specific to Computer Vision

vision Datasets, Transforms and Models specific to Computer Vision Installation First install the nightly version of OneFlow python3 -m pip install on

OneFlow 68 Dec 07, 2022
Geometry-Free View Synthesis: Transformers and no 3D Priors

Geometry-Free View Synthesis: Transformers and no 3D Priors Geometry-Free View Synthesis: Transformers and no 3D Priors Robin Rombach*, Patrick Esser*

CompVis Heidelberg 293 Dec 22, 2022
Part-Aware Data Augmentation for 3D Object Detection in Point Cloud

Part-Aware Data Augmentation for 3D Object Detection in Point Cloud This repository contains a reference implementation of our Part-Aware Data Augment

Jaeseok Choi 62 Jan 03, 2023
Perception-aware multi-sensor fusion for 3D LiDAR semantic segmentation (ICCV 2021)

Perception-Aware Multi-Sensor Fusion for 3D LiDAR Semantic Segmentation (ICCV 2021) [中文|EN] 概述 本工作主要探索一种高效的多传感器(激光雷达和摄像头)融合点云语义分割方法。现有的多传感器融合方法主要将点云投影

ICE 126 Dec 30, 2022
A High-Performance Distributed Library for Large-Scale Bundle Adjustment

MegBA: A High-Performance and Distributed Library for Large-Scale Bundle Adjustment This repo contains an official implementation of MegBA. MegBA is a

旷视研究院 3D 组 336 Dec 27, 2022
YoloAll is a collection of yolo all versions. you you use YoloAll to test yolov3/yolov5/yolox/yolo_fastest

官方讨论群 QQ群:552703875 微信群:15158106211(先加作者微信,再邀请入群) YoloAll项目简介 YoloAll是一个将当前主流Yolo版本集成到同一个UI界面下的推理预测工具。可以迅速切换不同的yolo版本,并且可以针对图片,视频,摄像头码流进行实时推理,可以很方便,直观

DL-Practise 244 Jan 01, 2023
N-Person-Check-Checker-Splitter - A calculator app use to divide checks

N-Person-Check-Checker-Splitter This is my from-scratch programmed calculator ap

2 Feb 15, 2022
StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.

3k Jan 08, 2023
Existing Literature about Machine Unlearning

Machine Unlearning Papers 2021 Brophy and Lowd. Machine Unlearning for Random Forests. In ICML 2021. Bourtoule et al. Machine Unlearning. In IEEE Symp

Jonathan Brophy 213 Jan 08, 2023
Code for "Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis

HAABSAStar Code for "Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis". This project builds on the code from https://gith

1 Sep 14, 2020
Official code for the paper: Deep Graph Matching under Quadratic Constraint (CVPR 2021)

QC-DGM This is the official PyTorch implementation and models for our CVPR 2021 paper: Deep Graph Matching under Quadratic Constraint. It also contain

Quankai Gao 55 Nov 14, 2022
A Neural Net Training Interface on TensorFlow, with focus on speed + flexibility

Tensorpack is a neural network training interface based on TensorFlow. Features: It's Yet Another TF high-level API, with speed, and flexibility built

Tensorpack 6.2k Jan 01, 2023
Python package for missing-data imputation with deep learning

MIDASpy Overview MIDASpy is a Python package for multiply imputing missing data using deep learning methods. The MIDASpy algorithm offers significant

MIDASverse 77 Dec 03, 2022
GuideDog is an AI/ML-based mobile app designed to assist the lives of the visually impaired, 100% voice-controlled

Guidedog Authors: Kyuhee Jo, Steven Gunarso, Jacky Wang, Raghav Sharma GuideDog is an AI/ML-based mobile app designed to assist the lives of the visua

Kyuhee Jo 5 Nov 24, 2021