curobo tutorial by using ur5e

This repository provides a environment for trying cumotion movit plugin with ur5e simulation. Although a docker image 'wuisky/isaac_ros_dev-x86_64:latest' is prepared for this tutorial, you can build by your self with origin build script like follows if you are using x86_64 cpu:

./build_image_layers.sh  --image_key x86_64.ros2_humble.user --image_name isaac_ros_dev-x86_64

Dependency

1. OS: Ubuntu 22.04 recommend(Jammy Jellyfish)
2. GPU and Memory: Ampere or higher NVIDIA GPU Architecture with 4 GB RAM or higher
3. NVIDIA driver
4. Docker and docker compose
5. nvidia-container-toolkit
6. vcstool

Quick start

After cloning this repository, enter the directory of this repository and execute the following command:

mkdir -p isaac_ros-dev/src;vcs import --recursive isaac_ros-dev/src < .rosinstall

1. docker compose up -d
2. Enter the container with docker exec -it isaac_ros_dev-x86_64-container bash, then run colcon build and source install/setup.bash

start ros2 nodes

Run the following commands in three terminals inside the container environment:

xacro -o ur5e.urdf /opt/ros/humble/share/ur_description/urdf/ur.urdf.xacro ur_type:=ur5e name:=ur5e;ros2 run isaac_ros_cumotion cumotion_planner_node --ros-args -p robot:=ur5e_preset_tool0_sphere.xrdf -p urdf_path:=$PWD/ur5e.urdf -p publish_robot_as_spheres:=true

this will cost for a while.After it shows "cuMotion is ready for planning queries!" you can proceed to next steps:

ros2 launch ur_robot_driver ur_control.launch.py ur_type:=ur5e robot_ip:=yyy.yyy.yyy.yyy initial_joint_controller:=joint_trajectory_controller use_fake_hardware:=true launch_rviz:=false

ros2 launch isaac_ros_cumotion_examples ur.launch.py ur_type:=ur5e use_fake_hardware:=true

you will see rviz is launched.

try motion planning

1. In MotionPlainng GUI, select Scene Objects tab and press Import button, navigate to workspaces/tutorial_scene.scene and open it.
2. Check hand under Current Scene Objects and attach it to tool0
3. Add MarkerArray visualization and set topics to /curobo/robot/spheres. This will display collision model used in curobo motion planner later.
4. Drag the interaction marker to the target pose and press Plan & Execute button. Then click Yes to publish planning scene.
5. You will see the collison model displayed in spheres and the robot is moving if the target pose is feasible.

Notes

In this tutorial, wu use a custom isaac_ros_cumotion branch, which is differ from upstream repository in follow aspects:

1. Take account for the attached object when executing path planning. This feature is reported in this PR
2. Add a ros param publish_robot_as_spheres for toggling publishing the robot sphere model as MarkerArray which is used in cumotion planner.
3. Add ur5e_preset_tool0_sphere.xrdf model file. 100 spheres are preseted in tool0 of ur5e model in this file for presenting attached object in cumotion planner.

Reminder

The hand model is downloaded from Thingiverse, and is licensed under the Creative Commons - Attribution - Non-Commercial - Share Alike license.

Although this tutorial is done by gui, you can achieve same operation by move group api which is more practical when developing your robot system. This tutorial just provides a "glimpse" for using cumotion moveit path planning plugin.