Robotics Manipulation Project

A comprehensive ROS 2 manipulation system for the UR3E collaborative robot with integrated object detection, motion planning, and gripper control capabilities.

Project Overview

This project provides a complete manipulation pipeline for robotic pick-and-place operations, combining:

• Motion Planning: MoveIt2-based trajectory planning and execution
• Object Detection: Point cloud-based surface and object detection
• Gripper Control: Robotiq 85 gripper manipulation with width/radian control
• Simulation & Real Hardware: Support for both Gazebo simulation and real UR3E robot

Project Structure

manipulation_project/
├── moveit2_scripts/          # Core manipulation executables (C++)
├── object_detection/         # Point cloud detection system (Python)
├── detection_interfaces/     # Custom ROS 2 message definitions
├── my_moveit_config/         # MoveIt2 configuration for simulation
├── real_moveit_config/       # MoveIt2 configuration for real robot
└── README.md                 # This file

Key Components

1. moveit2_scripts - Motion Planning & Control

C++ executables for robot manipulation:

• pick_and_place: Hardcoded pick-and-place sequence
• pick_and_place_perception: Vision-guided pick-and-place
• cartesian_move: Sequential Cartesian movements (Z then XY)
• gripper_control: Gripper control with width/radian modes
• move_joints: Direct joint angle control
• move_to_pose: Move to target end-effector pose

2. object_detection - Vision System

Python-based point cloud processing:

• Plane segmentation (table surface detection)
• Euclidean clustering (object detection)
• Marker visualization in RViz
• Custom message publishing (DetectedObjects, DetectedSurfaces)

3. detection_interfaces - Message Definitions

Custom ROS 2 message types:

• DetectedObjects: Object position, dimensions, ID
• DetectedSurfaces: Surface position, dimensions, ID

4. MoveIt2 Configurations

• my_moveit_config: Simulation environment (Gazebo)
• real_moveit_config: Real UR3E robot configuration

Quick Start

Build the Project

colcon build
source install/setup.bash

Run Simulation

# Terminal 1: Start MoveIt2 with Gazebo
ros2 launch my_moveit_config demo.launch.py

# Terminal 2: Run pick-and-place
ros2 run moveit2_scripts pick_and_place

Run with Object Detection

# Start MoveIt2 with object detection and RViz
ros2 launch moveit2_scripts pick_and_place_perception.launch.py

Control Gripper

# Open gripper (0 radians)
ros2 run moveit2_scripts gripper_control 0.0 radians

# Close gripper (0.8 radians)
ros2 run moveit2_scripts gripper_control 0.8 radians

# Gripper for 30mm object (width mode)
ros2 run moveit2_scripts gripper_control 0.03 width

Move Robot

# Cartesian move to position (meters)
ros2 run moveit2_scripts cartesian_move --ros-args -p x:=0.3 -p y:=0.0 -p z:=0.2

# Move to joint angles (radians)
ros2 run moveit2_scripts move_joints "0.0,-1.5708,0.0,-1.5708,0.0,0.0"

Key Features

• Sequential Cartesian Movements: Z-axis first, then XY plane (better control)
• Meter-based Units: All measurements in meters (not mm/cm)
• Dual Mode Gripper Control: Radians or object width
• Real-time Visualization: RViz integration with markers
• Perception-guided Manipulation: Automatic object detection and grasping

Dependencies

• ROS 2 (Humble or later)
• MoveIt2
• PCL (Point Cloud Library)
• UR3E URDF/description packages
• Gazebo (for simulation)

Documentation

See individual package READMEs for detailed information:

• moveit2_scripts/README.md
• object_detection/README.md
• detection_interfaces/README.md

Author

Dmitri Manajev (dmitri.manajev@protonmail.com)

License

This project is licensed under the MIT License - see the LICENSE file for details.