Upper-level-task-planning-of-Jaco-based-on-behavior-tree

Use behavior tree and DMP for upper-level task planning.

Background
Install
- Environment
- Step1
- Step2
- Step3
Usage
Result
Contributing

Background

This repository is a Kinova Jaco2 robot arm upper-level task planning Simulation Environment based on DMP and behavior tree.

Install

This project was developed based on ROS Melodic under Ubuntu 18.04, and the simulation environment was improved based on @freeman-1995 project JCAR-Competition. Based on his project, the Kinova Jaco v3 robot arm model was replaced with the Kinova Jaco v2 model.

Environment

This Package is tested on ubuntu18.04, ros melodic with GPU RTX 2060.

Step1

System environment : source code installation pointcloud library pcl with version1.8.1,vtk with version7.1.1 and opencv with version 4.1.2.

You need to add GPU options when compiling the above three packages.

Step2

The package find-object is used for object recognition.

Step3

Use the behavior tree framework to plan the tasks of the segmented dynamic motion primitives.

Usage

Compile the package

Clone the repository.

git clone https://github.com/Lygggggg/Upper-level-task-planning-of-Jaco-based-on-behavior-tree
cd Upper-level-task-planning-of-Jaco-based-on-behavior-tree

Go to the package named JCAR-Competition-master/robot_grasp, find the CmakeLists file inside. Comment out all the add_execytable and target_link.
catkin_make the workingspace.

cd Upper-level-task-planning-of-Jaco-based-on-behavior-tree
catkin_make

recover the items that are commented out before.
catkin_make the workingspace angin.

catkin_make

Run

Start the Gazebo simulation environment.

roslaunch robot_grasp START_Gazebo.launch

Start Rviz.

roslaunch robot_grasp START_Moveit.launch

Load the object to be grabbed into the Gazebo simulation environment.

$ roslaunch robot_grasp load_aruco_object.launch/loadall_object_part1.launch
# load_aruco_object.launch : Indicates that the QR code box is loaded;
# loadall_object_part1.launch : Indicates that some objects are loaded.

Start the find-object function package for object recognition, and select the object to obtain the position of the object.

roslaunch find_object_2d find_object_3d_simu.launch

Send the position of the identified object to the Ros parameter server.

rosrun robot_grasp find_and_pub

Start the behavior tree for object pick and place.

roslaunch behavior_tree_leaves pick_place_tree.launch
# or
roslaunch behavior_tree_leaves new_tree_like_guyue.launch

Result

Simulation experiment video.

Contributor

This project exists thanks to all the people who contribute.

Lygggggg/Upper-level-task-planning-of-Jaco-based-on-behavior-tree