vdbfusion_mapping: A C++ repository from Kin-Zhang

VDBFUSION_ROS_MAPPING

This repo is modified ROS1 version of vdbfusion for mapping incrementally based on received odometry and corresponding point cloud message. Since the origin repo of vdbfusion_ros have some problems on the cow dataset, more issues can be found here. The whole process is based on the ROS1, please check origin repo of vdbfusion if you'd like to use directly without ROS. Please remember clone submodules also.

# 内地的同学用gitee 快一点
git clone --recurse-submodules https://gitee.com/kin_zhang/vdbfusion_mapping.git
# OR from github
git clone --recurse-submodules https://github.com/Kin-Zhang/vdbfusion_mapping.git

Install

Docker version for convenient usage. [在内地的同学建议使用docker pull 先换一下dockerhub的源]

docker pull zhangkin/vdbmapping_mapping

# or build through Dockerfile
docker build -t zhangkin/vdbfusion_mapping .

# =========== RUN -v is the bag path in your computer
docker run -it --net=host -v /dev/shm:/dev/shm -v /home/kin/bags:/workspace/data --name vdbfusion_mapping zhangkin/vdbfusion_mapping /bin/zsh

Dependencies

Own environment, please check the file here, TESTED SYSTEM: Ubuntu 18.04 and Ubuntu 20.04. Here are some dependencies for desktop installed if you'd like to try. Please follow their dependencies to install, Dockerfile may help you with that also.

IGL: mesh save
OpenVDB: vdb data structure, ATTENTION Boost need 1.70, Ubuntu 18.04 default is 1.65
glog, gflag: for output log
ROS1: ROS-full (tested on melodic)

Usage

Please note that this is the for incremental mapping, no! odom output! So, you have to have odom/tf topic with same timestamp lidar msg. If you don't have the package to do so, checkout here: Kin-Zhang/simple_ndt_slam Really easy to get poses!! (But it didn't work well on depth sensor point cloud like cow dataset)

Config

The only thing you have to change is the config file about the topic name on your own dataset/equipment.

# input topic name setting ===========> Please change according to your dataset
lidar_topic: "/odom_lidar"
odom_topic: "/auto_odom"

# or tf topic ==> like the cow and lady dataset

Run

run launch with bag directly

roslaunch vdbfusion_ros vdbfusion_mapping.launch

save and pub map, open with visualization tools example image, .pcd file and .ply file (mesh)

rosservice call /save_map '/workspace/data/test' 0.0

Results

My own dataset with Kin-Zhang/simple_ndt_slam give the pose and this repo to mapping

ETH Zurich ASL: Cow and Lady RGBD Dataset Meshlab view, If the bag have the RGB info in the msg like XYZRGB etc, the results could be like right one:

Another dataset: TU Munich RGB-D SLAM Dataset and Benchmark - FR1DESK2 Test bag: rgbd_dataset_freiburg3_long_office_household-2hz-with-pointclouds.bag

Acknowledgement

ethz-asl/voxblox
PRBonn/vdbfusion
PRBonn/vdbfusion_ros
jianhao jiao: for the first version on vdbfusion mapping ros
paucarre: for the rgb version on vdbfusion

Style Formate: https://github.com/ethz-asl/linter

cd $YOUR_REPO
init_linter_git_hooks # install
linter_check_all # run

init_linter_git_hooks --remove # remove

Others

includes some todo and issue

grouped process TODO from voxblox
Memory increases a lot, maybe buffer and thread is not so great option, throw some data away if the queue is too long?
speedup again?? GPU? I don't know, let's find out

Solved:

For HDDA, here are comments from vdbfusion author @nachovizzo:

Additionally, the HDDA makes not much sense in our mapping context since it's highly effective ONCE the VDB topology has been already created