voxblox+orb_slam2+d435i

目录

 0. 环境配置

 1. 建立地图

0. 环境配置

    我的电脑是ubuntu16.04， ros版本是kinetic，传感器采用的是深度相机realsense D435i。

    安装voxblox可以按照github上的步骤，我是安装的voxblox-plusplus，里面有用到voxblox库。

   voxblox是不提供定位，所以需要额外系统提供定位。按照作者的说法

To work Voxblox needs a system to provide it with three pieces of of information

1.A 3D pointcloud of the environment, this will be remapped to voxblox's pointcloud topic. There are quite a few ways to get this, we have had success using

     - Depth cameras such as the realsense or Kinetic
     - Stereo cameras such as the VI-sensor, combined with a dense stereo pipeline           (see KITTI-dataset example).
     - 3D lidars such as made by Velodyne or Ouster

2.An estimate for the 6 degree of freedom pose of the system. This can be given using the ros tf tree (default option). In this case voxblox will look for a transform from the world frame to the frame the above point-cloud is given in. A seperate ros node is required to give this transform. Here there are also many options, we have previously used

     - Vicon / optitrack systems
     - Visual slam solutions such as ORB-SLAM
     - Visual-inertial solutions such as ROVIOLI, ROVIO or VINS-Mono.
     - Lidar ICP solutions
3.The transformation from the output of 2) to the frame the pointcloud in 1) is given in. If the same sensor is used for both this can just take the default value of identity.

1. 建立地图

  realsense可以输出直接输出points，上面第1个条件解决了。

  定位：采用orb_slam2， 可以直接安装ros包。

  第3个条件说的是如果位姿与点云坐标系是用的同一个传感器，那么就不需要提供两者之间的转换。

在voxblox_ros/launch文件夹下，新建一个launch文件夹：

如果use_tf_transforms设置的是true的话，下面这几个参数就无关紧要，不会被使用到。