Omnid Simulator and Moveit! Packages

This is the main repository of an Omnid group simulator and its Moveit! motion planning pipeline. If you haven't done so, check out my blog post for some higher level implementation detals.

There are 3 Delta robots and one object platform in this project.

The simulator includes:

• Robot model, including URDF and PyBullet model
• Joint control and joint information output
• Torsion springs mounted on each robot as serial elastic actuators
• Validation of torsion springs.

The motion planning pipeline includes

• omnid_moveit_config: a properly configured Moveit! configuration package
• Robot URDF model: URDF specifically for Moveit!
• Plugins:
  • omnid_move_group_interface: updates robot poses when the object's interactive marker is clicked and dragged
  • omnid_kinematics_plugin: IK plugin for a single delta robot
  • omnid_planning_adapter_plugin: post-processing planner adapter that generates the right trajectory waypoints

• Supplementary Packages:

  • control_msgs: this is the ROS control_msgs package, but as of Dec.2020 it is not available in Ubuntu 20.04 apt store yet.
  • tf2_armadillo: transformations for common Armadillo matrices and transform datatypes in ROS tf2 and geometry_msgs
    

  There's also some files for setting up the docker in docker_setup

  • docker file: docker image file
  • dockint: tool to build a docker container. Created by Dr. Matt Elwin
  • ros_settings.bash: settings to source when starting the docker file

Build the Package

The repo runs on Ubuntu Linux 20.04, ROS Noetic, rosdep, catkin_tools and several other prerequisites. As a universal way to setup the workspace,

1. Create workspace

       mkdir -p omnid/src
       cd omnid/src
       git clone https://github.com/RicoJia/Omnid_Project.git

2. Pull Dependencies

   • cd Omnid_Project Go to the project directory
   • Install VCS tool, see here
   • Use VCS tool to download all necessary packages. If VCS tool does not work, try git clone all pacakges in docker_setup/omnid_docs.repos manually.

        vcs import < docker_setup/omnid_docs.repos

   • cd ../.. go to the root of the package

3. Build a docker container and start it(Please use this dockerfile as it contains the latest dependencies we need)

   • Build the image

         cp src/Omnid_Project/docker_setup/dockint .
         cp src/Omnid_Project/docker_setup/Dockerfile .
         ./dockint from omnid $(pwd)/src/Omnid_Project/docker_setup

   • Build a container. Thanks Dr.Matt Elwin for sharing his dockint tool!

     ./dockint start omnid $(pwd)
     ./dockint run omnid bash

4. Build the package in the docker container

   • source /opt/ros/noetic/setup.bash this will setup some initial settings
   • catkin build build this package
   • source devel/setup.bash Source the workspace
   • source src/Omnid_Project/docker_setup/ros_settings.bash Source some correct ros settings that overwrite the previous ones
   • roslaunch omnid omnid.launch Launch the project
   • Have fun planning!

5. Once done with the project

   • exit to exit the container
   • ./dockint stop omnid to stop the container

Basic Usage

On Rviz,

1. Click and drag the marker embedded in the object platform to a desired pose
   • If we first start at unachievable poses, we can drag the marker anywhere we want until we reach the first achievable poses.
   • If we try to reach a valid pose to an unachievable pose, the marker will jump back to the closest valid pose
2. Hit plan to see the plan
3. Hit plan and execute to execute the plan
   • The simulator will return fail only if the action request from Moveit! is preempted.
     This is because Moveit! checks the proximity between the plan and the actual robot TF. Also it checks if the controller takes too long to execute a path. see move_group/trajectory_execution/execution_duration_monitoring

External Resources

1. Notes for Developers
2. Blog Post
3. Video Demonstration