Skip to content

Latest commit

 

History

History
139 lines (108 loc) · 3.83 KB

README.md

File metadata and controls

139 lines (108 loc) · 3.83 KB

ur3_edge_follower

System Preparation

A. Realsense Installation

  1. Install realsense2-camera for ROS Melodic
sudo apt install ros-$ROS_DISTRO-librealsense2 ros-$ROS_DISTRO-realsense2-camera ros-$ROS_DISTRO-realsense2-description
  1. Copy the udev file for realsense-device
wget https://github.com/IntelRealSense/librealsense/raw/master/config/99-realsense-libusb.rules
sudo cp 99-realsense-libusb.rules /etc/udev/rules.d/

B. JSK PCL Installation

  1. Install the core JSK-PCL package
sudo apt-get install ros-melodic-jsk-pcl-ros
  1. Allow the system to check the normal vector
sudo apt-get install ros-melodic-jsk-rviz-plugins
  1. Provide organized pointcloud from the camera
sudo apt-get install ros-melodic-rgbd-launch
  1. Install ros-numpy
sudo apt-get install ros-melodic-ros-numpy
  1. Install python-pcl
pip install python_pcl-0.3.0rc1-cp27-cp27mu-linux_x86_64

C.Camera Bringup

You can test the realsense camera and jsk pcl package by the following instructions.

  1. This invoke the camera to produce Organized point clouds
roslaunch realsense2_camera rs_rgbd.launch camera:=d415
  1. This will show you the RViz with all the different edges detected
roslaunch jsk_pcl_ros sample_organized_edge_detector.launch

You may encounter an error about fail to load a .bag file. You need to go to /opt/ros/melodic/share/jsk_pcl_ros/sample/sample_organzied_edge_detector.launch and comment out line 4 to fix this issue. That line is about to load a play_rosbag_shelf_bin.xml.

  1. Edit the launch file by changing the input topic to your camera's pointcloud.
# line 13 and line 14
<remap from="~input" to="/d415/depth_registered/points"/>
# YOu need to change this topic if you are using other camera.

Robot Vision Control

  1. Bringup the robot
./start_robot_bringup.sh
  1. Start robot pcl planning rviz
./start_robot_pcl_planning.sh
  1. Start edge following
rosrun edge_follwer main.py

image

Pointcloud path planning

Reference: https://github.com/vincent51689453/pointcloud_path_planning

cd src/path_analyze
python3 smoothing.py

Input pointcloud

image

Planned path

image

Basic Demonstration

image

image

Information

A. ROS related

In order to create a C++ PCL package, you need to perform several actions.

  1. Create a ros package
# You have to replace your own package name with my_pcl_tutorial
catkin_create_pkg my_pcl_tutorial pcl_conversions pcl_ros roscpp sensor_msgs
  1. Modify package.xml and add the followings
<build_depend>libpcl-all-dev</build_depend>
<exec_depend>libpcl-all</exec_depend>
  1. Modify CMakeLists.txt of the package
# You have to replace your xxx.cpp with pcl_subscribe.cpp
add_executable(${PROJECT_NAME}_node src/pcl_subscribe.cpp)
target_link_libraries(${PROJECT_NAME}_node ${catkin_LIBRARIES})

  1. VSCode configuration. Sometimes your vscode may not able seek for the ros header file. Therefore you can use my .vscode/c_cpp_properties.json to solve this issue.

  2. Execute a C++ node

# <executable_name> may not be the name of .cpp file.
rosrun <package_name> <executable_name>

B. Reference

  1. jsk_pcl_ros/sample/sample_organized_edge_detector.launch

  2. jsk_pcl_ros_utils/sample/sample_normal_concatenater.launch

  3. jsk_pcl_ros/sample/sample_normal_estimation_omp.launch

  4. jsk_pcl_ros/sample/sample_line_segment_detector.launch