SRL_GraspableTargetDetection

Geometry based graspable target detection for legged robots.

Originated in the Space Robotics Lab (SRL) of the Tohoku University. Based on ClimbLab (Uno et al. 2022).

Authors and Maintainers

• Jitao Zheng ([email protected])

• Taku Okawara ([email protected])

• Kentaro Uno ([email protected])

• Antonin Ringeval-Meusnier ([email protected])

System Requirements

The code was tested on:

• ROS Humble
• Ubuntu 22.04
• gcc version 11.4.0

You will require the following packages and libraries:

• Point Cloud Library (PCL)
• LibInterpolate
• Eigen3

Environment Setup

1. Installation of Catch2

   git clone https://github.com/catchorg/Catch2.git
   cd Catch2
   cmake -B build -S . -DBUILD_TESTING=OFF
   sudo cmake --build build/ --target install

2. Installation of LibInterpolate

   git clone https://github.com/CD3/libInterpolate
   cd libInterpolate
   mkdir build
   cd build
   cmake ..
   make
   sudo make install

3. Make Sure that you have ALL the pcl libraries installed

   If you don't have yet installed Point Cloud Library (PCL), you can install it by typing:

   sudo apt install libpcl-dev
   sudo apt install ros-humble-pcl-ros

Quick Start

Open a new terminal window.

Type:

cd ~/ros2_ws/src

Clone the repository:

git clone [email protected]:Space-Robotics-Laboratory/graspable_points_detection_ros2.git

Then we can build it:

cd ..
colcon build

IMPORTANT, don't forget to sourcing the packages you just build in order for them to show when you want to run them:

source install/setup.bash

We will first test the algorithm on some examples. You will need at least three terminals.

1. Terminal 1

   Launch the graspable target detection.

   cd ~/ros2_ws
   source install/setup.bash
   ros2 run detect_graspable_points detect_graspable_points_main

   The algorithm subscribes to a point cloud message merged_pcd in sensor_msgs/PointCloud2 format. So in principle, you can subscribe to any tope

2. Terminal 2

   Publish stored example point cloud.

   cd ~/ros2_ws
   source install/setup.bash
   ros2 run detect_graspable_points publish_pointcloud2

   Now, a point cloud in .pcd format will be published once per second as sensor_msgs/PointCloud2 message under the topic /merged_pcd in the regression_plane_frame coordinate frame.

3. Terminal 3

   Open RViz.

   cd
   ros2 run rviz2 rviz2

   You can freely choose which topic you want to visualize, whether that is the raw point cloud with the graspable points (light green spheres) or the color gradient of Graspability Score.

Input Point Cloud

Copy the point cloud you want to examine into the folder detect_graspable_points/src/pcd_data/ and change the path in the program publish_pointcloud2.cpp. Make sure that it is in .pcd format. Of course the algorithm can also subscribe to any PointCloud2 message with the name /merged_pcd coming from other nodes.

Output

This package output several data for visualization. If you are only interested in the graspable point it is pulished as /detect_graspable_point/graspable_points. To choose the different gripper in this file just turn to true the gripper you want to use and flase the others.

Gripper Parameters

The edition of parameters now takes place in settings.hpp in include/detect_graspbable_points folder.

Orientate yourself to the simplified dimensions of this gripper:

No.	Variable	Explanation
1	Palm diameter	Diameter of gripper's palm
2	Palm diameter of finger joints	Distance between two opposite first finger joints
3	Finger length	Length of the first finger segment
4	Spine length	Length of the last finger segment
5	Spine depth	Length of the spine itself
6	Opening angle	Maximum opening angle
7	Closing angle	Maximum closing angle
8	Opening spine radius	Distance from the center of the palm to the tip of the furthest spine
9	Opening spine depth	Distance from the horizontal plane to the tip of the spine when opened
10	Closing height	Vertical distance between the tip of the spine and the bottom of the palm when closed

TODO: I don't know what exactly "margin_of_top_solid_diameter" and "inside_margin_of_bottom_void_diameter" are. Please complement this Readme if anyome knows more.

Matching Parameters

The matching parameters are the most essential parameters for the graspability and curvature analysis.

No.	Variable	Data type	Explanation
1	Voxel size	float [m]	Size of voxels you want to use
2	Threshold of Solid Voxels	int	Number of solid voxels needed to not apply penalty coefficient
3	Artificial_add_point	String (on/off)	True if you think your map is too sparse for running properly
4	Delete lower targets z-threshold	float [m]	Threshold below which point are not checked for graspability
5	Auxiliary void voxel layers above gripper mask	int	Empty layers to avoid false matching on too convex shape
6	Graspability threshold	int	Graspability above which we consider graspable with certainty

Troubleshooting

• If libinterpolated dosn't get build even after installing catch2, make sure to have GSL installed and restarting your computer might help.