MoveIt 2 Panda Learning Demo

English | 简体中文

A lightweight MoveIt 2 learning project using the Franka Panda robot in RViz. It demonstrates joint-space goal planning, pose goal planning, PlanningScene collision objects, and OMPL obstacle avoidance on Ubuntu 24.04, ROS 2 Jazzy, MoveIt 2, RViz2, and ros2_control fake hardware.

This repository is a learning / skill-validation demo rather than a custom production MoveIt 2 stack.

The repository contains only the learning/demo code built on top of ROS 2, MoveIt 2, and the official Panda resources. MoveIt, ROS 2, and Panda resources are provided by their respective projects.

Demo

The main visual demo is obstacle avoidance with a PlanningScene collision object:

Additional recordings are stored in media/.

Features

1. Joint Goal Planning
   Reads the current 7 Panda arm joints, changes panda_joint1 by about +0.3 rad, plans with OMPL, and executes the trajectory.

2. Pose Goal + IK
   Reads the current end-effector pose, moves the target pose upward by about 5 cm, lets MoveIt solve the kinematics/IK goal, then plans and executes.

3. PlanningScene + Obstacle Avoidance
   Adds a demo_box collision object to the PlanningScene, plans a collision-free motion with OMPL, executes it, and removes the object before exit.

System Architecture

flowchart TD
    A["Application Node"] --> B["MoveGroupInterface"]
    B --> C["/move_group"]
    C --> D["Robot Model / IK / PlanningScene"]
    D --> E["OMPL"]
    E --> F["Trajectory"]
    F --> G["ros2_control"]
    G --> H["Panda fake hardware in RViz"]

Loading

For more detail, see docs/architecture.md.

Additional Demo Recordings

Joint Goal

MP4

joint_goal_demo shows direct joint-space planning where the target joint vector is known.

Pose Goal

MP4

pose_goal_demo shows how a Cartesian end-effector pose target is converted by MoveIt into a valid robot state before motion planning.

Obstacle Avoidance

MP4

planning_scene_obstacle_demo is the main visual demo: a collision box is added to the PlanningScene and OMPL searches for a collision-free path around it.

Overview

MP4

Key Concepts

MoveGroupInterface

The high-level C++ API used by application nodes to set goals, plan trajectories, and request execution.

move_group

The core MoveIt ROS 2 backend node. It loads the robot model, monitors state, handles planning requests, manages the PlanningScene, and coordinates trajectory execution.

Joint Goal

A joint goal directly specifies q_goal, the desired robot joint values. No end-effector pose-to-IK step is needed.

Pose Goal

A pose goal specifies an end-effector pose. MoveIt uses kinematics/IK to find a legal joint state that satisfies the pose before planning.

PlanningScene

MoveIt's internal world model containing the robot state, collision objects, and information needed for collision checking.

OMPL

The Open Motion Planning Library provides sampling-based planners that search for collision-free paths in configuration space.

MoveIt vs DLS IK

Damped Least Squares IK is local differential kinematics:

Pose error -> Jacobian -> delta q

It is direct, interpretable, and useful for local continuous control or high-frequency online correction.

MoveIt + OMPL is global motion planning:

start -> collision-aware search -> goal

It can use the PlanningScene, collision checking, and sampling-based search for global obstacle avoidance. They are complementary rather than replacements: MoveIt/OMPL can handle global collision-free motion, while Jacobian/DLS methods are suitable for local task-space control and online correction.

Requirements

• Ubuntu 24.04
• ROS 2 Jazzy
• MoveIt 2
• RViz2
• ros2_control and ros2_controllers
• Official MoveIt Panda resources

This project was validated on WSL2 + WSLg, but it is a standard ROS 2 workspace and is not limited to WSL.

Installation

Install the ROS 2 / MoveIt packages used by the demos:

sudo apt update
sudo apt install ros-jazzy-moveit
sudo apt install ros-jazzy-moveit-resources-panda-moveit-config
sudo apt install ros-jazzy-ros2-control ros-jazzy-ros2-controllers

Build

Clone the repository and build it as a ROS 2 workspace:

source /opt/ros/jazzy/setup.bash
colcon build --symlink-install
source install/setup.bash

Or use:

./scripts/build.sh

Run

Terminal 1:

source /opt/ros/jazzy/setup.bash
ros2 launch moveit_resources_panda_moveit_config demo.launch.py

Terminal 2:

source /opt/ros/jazzy/setup.bash
source install/setup.bash
ros2 run panda_moveit_learning joint_goal_demo

ros2 run panda_moveit_learning pose_goal_demo

ros2 run panda_moveit_learning planning_scene_obstacle_demo

Convenience scripts are also provided:

./scripts/launch_panda_demo.sh
./scripts/run_joint_goal_demo.sh
./scripts/run_pose_goal_demo.sh
./scripts/run_obstacle_demo.sh

Troubleshooting

If a vendor library such as libsdformat14.so.14 cannot be found, first confirm the ROS environment is sourced:

source /opt/ros/jazzy/setup.bash
echo "$LD_LIBRARY_PATH" | tr ':' '\n' | grep sdformat

Do not modify the official Panda URDF, SRDF, joint limits, controller configuration, kinematics settings, or self-collision matrix for these demos.

License

This repository is released under the MIT License. ROS 2, MoveIt 2, OMPL, and Panda resources are provided by their respective projects and licenses.