Graph SLAM

Environment

• Ubuntu16.04
• ROS(Kinetic)
• g2o
• PCL(1.8)

Setup

Graph SLAM

Download Graph SLAM package

(catkin workspace)
$ git clone https://github.com/Sadaku1993/graph_slam

g2o

Download g2o package

Requirements

• cmake
• Eigen3
• suitesparse
• Qt5
• libQGLViewer

On Ubuntu these dependencies are resolved by installing the following packages.

• cmake
• Eigen3
• libsuitesparse-dev
• qtdeclarative5-dev
• qt5-qmake
• libqglviewer-dev

Install

( graph_slam package )
$ ./setup.bash
$ cd Thirdparty
$ git clone https://github.com/Sadaku1993/g2o
$ cd g2o
$ mkdir build
$ cmake .. -DBUILD_CSPARSE=ON
$ make
$ sudo make install

Velodyne

Download Velodyne package

( catkin workspace )
$ git clone https://github.com/ros-drivers/velodyne

Normal Estimation

Download Normal Estimation Package

( catkin workspace )
$ git clone https://github.com/amslabtech/perfect_velodyne

Download Bag Data

This bagdata is recorded at Meiji University (Autonomous Mobile Systems Lab).

Move this bagdata to bagfiles dirs.

graph_slam/
  data/
    bagfiles/
      Infant-Dkan.bag

How to Use

1. Save Data
2. Remove Obstacle
3. GICP
4. Graph Optimization and Loop Closing

Save Data

save odometry and pointcloud data.

$ roslaunch graph_slam data_save.launch

odometry is saved at

graph_slam
    /data
        /csv/odometry.csv

pointcloud is saved at

/graph_slam
    /data
        /pcd
            /0.pcd
            /1.pcd
             :
             :
             :

Remove Obstacle (By Clustering)

in the pointcloud data, obstacles (like human and so on) are included. in order to removed them, use clustering(PCL)

$ roslaunch graph_slam remove_cluster.launch

pointcloud is saved at

/graph_slam
    /data
        /remove
            /0.pcd
            /1.pcd
             :
             :
             :

Gicp

$roslaunch graph_slam node_edge.launch

node edge data is saved at

/graph_slam
  /data
    /csv
      gicp.csv

Loop Closing

$roslaunch graph_slam slam3d.launch

update node edge data is saved at

/graph_slam
  /data
    /csv
      /0.csv
      /1.csv
       :
       :
       :

View Created 3D Map

prease serect csv file name and detect pcd file name

$roslaunch graph_slam integrator.launch

  CSV file list
    odometry.csv
    gicp.csv
    0.csv
    1.csv
    :
    :
  
  Please select csv file: 2.csv
  Please enter map file: 2.pcd
    :
    :