IITP Anthrophomorphic Robot Arm & Hand Project
Manipulator controlling ROS package with moveit! for V-rep simulation / Local kinova jaco2 machine
Currently tested in Ubuntu 18.04 and ROS melodic
Installation in Ubuntu 18.04 with ROS melodic is recommended (since other version of ROS or ubuntu distro were not tested)
V-rep source can be downloaded from here and required to be installed within the /opt folder. Installed location can be varied, but should be matched with the vrep_path argument within the launch file: vrep_jaco_bringup/launch/bringup.launch: vrep_path. If the extracted folder is named CoppeliaSim_Edu_V4_0_Ubuntu in ~/Downloads folder, then the vrep_path argument should be ~/Downloads/CoppeliaSim_Edu_V4_0_Ubuntu.
If have you install Coppelia (v4.0.0) instead of V-rep (v3.6.n), you need to move file:libsimExtROSInterface.so from folder:compiledRosPlugins to the source folder. If there is no such folder named compiledRosPlugins, then the below will not be of your case. It has already done for you. (This applies to v4.1.0 or later)
Within the source file:
cd compiledRosPlugins
mv libsimExtROSInterface.so ..Since we are not using official repo of moveit, it is required to create and build moveit from the modified source of our repo.
mkdir -p ~/name_of_your_workspace/moveit ~/name_of_your_workspace/vrep_jacoBecause vrep_jaco package requires moveit package to be overlayed, building and sourcing prior to making vrep_jaco is required. This folder will be later used to store the repo of moveit source and the vrep_jaco repo along side. Within your moveit folder in workspace,
cd ~/name_of_your_workspace/moveit
wstool init srcto create src folder with .rosinstall file in it.
It is assumed that the one is capable of installing ROS with full compatibility. This package is running python3, so proper steps to deal with python3 should be followed, Additionally, standalone catkin package is also required, so it should be installed along with the ROS default catkin_make
sudo apt-get install python3-pip python3-yaml
pip3 install rospkg catkin_pkg
sudo apt-get install python-catkin-tools // if using Ubuntu
pip install -U catkin_tools // if using other OSWithin your moveit folder, (the directory where your src folder is at)
wstool merge -t src https://raw.githubusercontent.com/TMmichi/moveit/master/vrep_jaco_moveit.rosinstall
wstool update -t src
rosdep install -y --from-paths src --ignore-src --rosdistro ${ROS_DISTRO}
catkin config --extend /opt/ros/${ROS_DISTRO} --cmake-args -DCMAKE_BUILD_TYPE=Release
sudo apt-get install libusb-dev libbluetooth-dev libcwiid1 libcwiid-dev
catkin buildSource setup.bash file in your project devel folder to the .bashrc.
source ~/name_of_your_workspace/moveit/devel/setup.bashWithin the moveit package, ompl package and ik algorithm from TRAC is used, and hence should be installed properly.
sudo apt-get install ros-${ROS_DISTRO}-ompl
sudo apt-get install ros-${ROS_DISTRO}-trac-ik-kinematics-pluginsudo apt-get install libqt5x11extras5=5.5.1-3build1in order to install ros-vrep_jaco_moveit.rosinstall.
(MANDATOROY: YOU MUST SOURCE setup.bash IN MOVEIT DEVEL FOLDER)
After sourcing the setup.bash file within moveit to overlay, go to your vrep_jaco folder in your workspace, and clone package.
cd ~/name_of_your_workspace/vrep_jaco
git clone https://github.com/TMmichi/vrep_jaco.git
mv vrep_jaco srcIt is required to have all of your sources in a folder name src, so please change vrep_jaco folder with the command above.
Build your repo and source it. With the following command, sourcing will done automatically within the terminal at launching.
catkin_make
echo "source ~/name_of_your_workspace/vrep_jaco/devel/setup.bash" >> ~/.bashrc
source ~/.bashrcSince we are teaking ROS timer with sim_time, it is required to modify timer.py in ROS package. Comment line 163, where it says raise rospy.exceptions.ROSTimeMobedBackwardsException(time_jump). This is necessary because we are constantly reseting our ROS timer at the reset call of each episodes. You can find timer.py file with the following command below.
cd /opt/ros/${ROS_DISTRO}/lib/python2.7/dist-packages/rospypip install -r requirements.txtManipulation of a real machine and one in the simulation are much alike from each other.
roslaunch vrep_jaco_bringup bringup.launchbringup.launch file will launch a V-rep env with the scene file jaco_table2.ttt
Bringup launch file DOES NOT include manipulator URDF xacro.
roslaunch rl_controller rl_controller.launchPrior to connect ROS-api of V-rep with the simulation environment itself, V-rep should be launched with the designated port number. Default port number has been set to 19997 in rl_controller/scripts/env/env_vrep_client.py line number 15 with server_port.
If the api client node is running on a separate machine other than the machine with V-rep simulation, IP address should also be clarified, other than the default localhost (127.0.0.1) in rl_controller/scripts/env/env_vrep_client.py line number 14 with server_address.
roslaunch jaco_controller_integrated jaco_controller_integrated.launchjaco_controller_integrated.launch includes most of the parameters required from the moveit! package with control parameters and launches the visualization node with RVIZ.
It also launches the ROS node with C++ script jaco_controller_integrated.cpp in folder jaco_controller_integrated/src which initializes the actionlib client side which communicates with the v-rep api server side.
By using Moveit! package, user does not have to consider action/state synchronization between the machine in the simulation and the controller node. Providing target gripper pose to the plan() method within the move_group instance will solve inverse kinematics problem for designated pose and store the solution as a member variable.
Calling sendgoal() method within the action_client will publish rostopic with name and type in /USER_DEFINED_NAME/follow_joint_trajectory/goal and control_msgs/FollowJointTrajectoryActionGoal respectively. Published topic from move_gropup instance in actionlib client side will then be transfered to the server side and wait till all of the goal joint states in follow_joint_trajectory/goal to be finished in the simulation.
roslaunch kinova_bringup kinova_robot.launchroslaunch jaco_controller_kinova jaco_controller_kinova.launchAs it is mentioned before, there is no significant differnece with the simulation ros control script.
Added: jaco_controller_integrated.launch will take over both of the simulation side and the real side in control. (WIP)
Consists:
- action_client
- jaco_arm_pkgs
- jaco_controller_integrated
- kinova-ros
- moveit
- rl_controller
- vrep_api
- vrep_jaco_bringup
- vrep_jaco_data
- vrep_jaco_moveitaction_client package wraps vrep environment with C++ client. It uses vrep_api package
jaco_arm_pkgs package provides URDF information of jaco robot for the moveit! to consider
jaco_controller_integrated package is a C++ vrep simulated controller script named 'jaco_controller.cpp' loacted in src folder. It uses moveit! for the low-level control. By providing target pose of the gripper on line 78 it planns its trajectory and velocity profile, and on line 84, it moves as its given profile.
kinova-ros package is a prerequisite for the real jaco.
moveit is the moveit! package, and shouldn't be eddited.
rl_controller is the main work scheme script, where it encodes states, produce action from policy, and trains via RL and IRL.
Should build ROS package for python3 in order to use moveit!. Follow instructions given in hyperlink: Instructions
kinova-ros_source_code from github@kinova-robotics
V-rep_api from github@JoshSong
controller_script from moveit_tutorial