🤖 LatteBot

Mobile Manipulation for Smart Farming

Autonomous dairy farm operations using industrial robotics and distributed systems

Overview • Features • Technologies • Architecture • Structure • Contributors

---

🎯 Overview

LatteBot is a ROS-based robotic system that automates dairy farm operations using a Universal Robots UR10e manipulator with Robotiq 2F-140 gripper mounted on a 10-meter linear mobile platform. The system demonstrates autonomous bucket handling and delivery to multiple cow stalls through a distributed MQTT/ROS architecture.

What It Does

graph LR
    A[Operator UI] -->|MQTT| B[Bridge]
    B -->|ROS| C[Robot Controller]
    C --> D[Platform Motion]
    C --> E[Arm Planning]
    C --> F[Gripper Control]
    D --> G[Gazebo Simulation]
    E --> G
    F --> G
    H[Arduino + Weight Sensor] -->|Serial| I[Serial Bridge]
    I -->|MQTT| B

Loading

System Capabilities:

• 🦾 6-DOF robotic arm with industrial parallel-jaw gripper
• 🚂 10-meter linear platform for multi-station operations
• 🌐 Multi-platform architecture (Linux + Windows coordination)
• 📡 MQTT/ROS bridge with cloud-based message broker
• 🎮 Interactive GUI for sequence planning and monitoring
• ⚖️ Arduino weight sensors for milk consumption monitoring

---

✨ Key Features

🎪 Complex Scenarios ✅ Incomplete feeding detection with timeout ✅ Parallel task management and queueing ✅ Multi-cow delivery sequences ✅ Automatic bucket lifecycle management

🔧 Advanced Control ✅ Three-phase grasp sequences (100% reliability) ✅ Collision-aware motion planning ✅ Thread-safe state management ✅ Dynamic model spawning

---

🛠️ Interesting Techniques

🤖 ROS and Robotics

Technique	Description	File
URDF/Xacro Macros	Parametric robot descriptions with macro-based assembly	urdf/ur10e.urdf.xacro
MoveIt Planning	Collision-aware trajectory planning with named poses	ur10e_moveit_config/
Gazebo Physics	Custom contact parameters (μ=2000, kp=5M) for reliable grasping	models/bucket/model.sdf
ROS Control	Hardware abstraction for arm, gripper, and platform	controller/
TF Transforms	Complete kinematic chain from world to end effector	System-wide

🐍 Python Control Systems

• Thread-Safe Queues: Lock-based sequence execution prevents race conditions → bridge_keypad2robot.py
• State Machine Pattern: Multi-phase grasp (open → pre-grasp → close) → robot_movement.py
• Action Servers: Non-blocking trajectory execution with feedback
• Dynamic SDF Modification: Runtime model name changes for grasp plugin compatibility

📡 MQTT Integration

• Paho MQTT: TLS-encrypted pub/sub
• Bidirectional Bridging: MQTT ↔ ROS topic translation with state tracking
• Distributed Architecture: Windows UI ↔ Cloud Broker ↔ Linux Controller

🎨 GUI Development

• Tkinter: Custom numerical keypad with sequence planning → numerical_keypad.py
• Scrollable Canvas: Dynamic content with mousewheel support
• State Visualization: Real-time cow availability tracking

🔌 Arduino Integration

• Custom Binary Protocol: Efficient serial communication with header (\xff) and payload structure → bridge_serial2MQTT.py
• Weight Monitoring: Load cell simulation with button-based weight changes and LCD display
• Auto-detection: COM port discovery via device description matching
• Event-Based Publishing: Triggers MQTT messages when milk consumption thresholds are met or timeout occurs (20 seconds)
• Serial → MQTT Bridge: Translates Arduino sensor data to MQTT topics (cow/{calf_num}) for robot coordination

---

📚 Technologies and Libraries

🤖 Robotics Stack

• ROS Noetic - Robot Operating System for Linux/WSL
• MoveIt - Motion planning framework with OMPL/CHOMP planners
• Gazebo 11 - 3D physics simulator with ODE solver
• Universal Robots ROS Drivers - UR10e kinematics and controllers
• Robotiq Gripper Package - 2F-140 gripper URDF with mimic joint plugin
• Gazebo Grasp Plugin - Contact-based grasping simulation

📡 Communication

• HiveMQ Cloud - Managed MQTT broker with TLS encryption (port 8883)
• Paho MQTT Python - v1.6.1 client library
• PySerial - Arduino integration for weight sensor simulation

🐍 Python Environment

• rospy - ROS Python client library
• moveit_commander - Python interface for MoveIt
• tf.transformations - Quaternion and Euler angle conversions
• ConfigParser - INI file parsing for MQTT/serial settings

🔧 Hardware & Embedded

• Arduino - Microcontroller for weight sensor simulation
• Adafruit SSD1306 - OLED display library for weight visualization
• Custom Serial Protocol - Binary message format with header/payload/footer structure

---

📁 Project Structure

📦 lattebot/
├── 📂 .github/                       # GitHub configuration
│   ├── 📄 copilot-instructions.md    # Complete system architecture reference
│   └── 📄 COPILOT_INSTRUCTIONS_UPDATE.md
├── 📂 pkg01/                         # Main ROS package ⭐
│   ├── 📂 config/                    # MQTT/serial configuration
│   ├── 📂 controller/                # Joint controller parameters
│   ├── 📂 launch/                    # ROS launch files for simulation
│   ├── 📂 models/                    # Custom Gazebo models (bucket, cow)
│   ├── 📂 scripts/                   # Python control scripts & bridges
│   │   ├── 🐍 robot_movement.py      # Main robot controller
│   │   ├── 🐍 bridge_keypad2robot.py # MQTT → ROS bridge
│   │   ├── 🐍 bridge_serial2MQTT.py  # Arduino → MQTT bridge
│   │   ├── 🐍 pickup_site.py         # Operator interface
│   │   ├── 🐍 numerical_keypad.py    # GUI components
│   │   └── 📂 calf_arduino/          # Arduino firmware
│   │       └── 🔧 calf_arduino.ino   # Weight sensor code
│   ├── 📂 urdf/                      # Robot description (xacro format)
│   ├── 📂 world/                     # Gazebo world definitions
│   ├── 📂 claude_explanations/       # Troubleshooting documentation 📖
│   └── 📂 meshes/                    # 3D models for visualization
├── 📂 ur10e_moveit_config/           # MoveIt configuration package ⭐
│   ├── 📂 config/                    # Planning parameters & SRDF
│   └── 📂 launch/                    # Motion planning launch files
├── 📂 robotiq/                       # Robotiq gripper packages (gitignored)
├── 📂 roboticsgroup_gazebo_plugins/  # Mimic joint plugin (gitignored)
└── 📄 requirements.txt               # Python dependencies

📌 Key Directories

Directory	Description
pkg01/claude_explanations/	📖 Comprehensive markdown docs covering physics tuning, collision troubleshooting, and grasp reliability - essential reading
pkg01/scripts/	🐍 All executable Python scripts: robot controller, MQTT bridges, test utilities
pkg01/scripts/calf_arduino/	🔧 Arduino firmware for weight sensor simulation with OLED display
pkg01/models/bucket/	🪣 Custom SDF model with extreme physics (μ=2000, kp=5M) for stable grasping
ur10e_moveit_config/config/	⚙️ MoveIt configuration with manually tuned collision matrices and named poses

---

🏗️ System Architecture

The project uses a three-tier distributed architecture:

Component Overview

Component	Platform	Purpose
Operator Interface	Windows	Tkinter GUI for planning cow milking sequences
MQTT Bridge	Linux/WSL	Translates MQTT messages to ROS topics with queue management
Robot Controller	Linux/WSL	Orchestrates platform motion, MoveIt planning, and gripper control
Serial Bridge	Windows	Reads Arduino sensor data and publishes to MQTT
Arduino System	Hardware	Monitors bucket weight via load cells with OLED display
Gazebo Simulation	Linux/WSL	Executes physics-based movements with contact-based grasping

Communication Flow

MQTT Topics:

• Pickup-Site → Complete sequence data from operator to bridge
• cow/{calf_num} → Weight sensor triggers from Arduino to bridge

ROS Topics:

• /calf_num → Task commands from bridge to robot controller
• /ur10e_robot/joint_states → Joint positions for monitoring
• /ur10e_robot/*_controller/follow_joint_trajectory → Action servers for motion execution

Message Protocol: HiveMQ Cloud broker with QoS 2 (exactly-once delivery) ensures reliable communication across Windows/Linux boundary.

Arduino Serial Protocol

The Arduino communicates via a custom binary protocol over serial (9600 baud):

[ 0xFF | Calf_Num | Weight | 0xFE ]
  ^       1 byte    1 byte    ^
  Header                       Footer

Trigger Conditions:

• ✅ Weight drops below starting_weight - milk_limit → Publishes 1 (cow finished)
• ⏰ 120-second timeout → Publishes 0 (incomplete feeding)

---

Acknowledgements

This project builds upon several open-source robotics packages:

• roboticsgroup_gazebo_plugins - Mimic joint plugin for Robotiq gripper linkage simulation
• ros-industrial-attic/robotiq - Robotiq 2F-140 URDF models and visualization meshes
• gazebo-pkgs - Gazebo grasp fix plugin for contact-based object grasping
• general-message-pkgs - Supporting message definitions for grasp plugin

---

Contributors

Casali Cristian

Flotta Aldo

---

Developed as part of Smart Robotics course project, A.Y. 2024/2025