Smart Warehouse Robot

Autonomous warehouse navigation system based on ROS 2 Foxy. This project implements a distributed architecture where a generic ROS 2 robot handles intelligent navigation while an off-board host PC processes AI vision.

System Architecture

This project uses a split processing architecture to maximize the robot's hardware performance:

1. Robot: Handles Hardware Bringup, SLAM, Nav2 Pure Pursuit, HTTP Video Streaming, and the Mission State Machine.
2. Host PC: Connects to the robot's camera stream, runs YOLOv8 inference, and sends real-time obstacle detections back to the robot via a zero-lag UDP socket (Port 5005).

Requirements

Robot

• OS: Ubuntu 20.04 (Focal Fossa)
• ROS 2: Foxy Fitzroy (with nav2 and slam_toolbox packages)
• (Optional) Docker and NVIDIA Container Runtime.

Host PC

• Python: 3.8+
• Libraries: ultralytics, torch, opencv-python, numpy.
• A CUDA-capable GPU is highly recommended for real-time YOLOv8 inference.

Repository Structure

smart_warehouse_robot/
├── models/             # YOLOv8 Neural Network weights (e.g., best.pt)
├── scripts/            # Core Python nodes
│   ├── mission_vision.py    # Robot: UDP Receiver & Mission Controller
│   └── pc_vision_node.py    # Host PC: YOLOv8 Inference Node
├── src/                # Custom ROS 2 packages and Nav2 parameters
├── utils/              # Diagnostic scripts, fixes, and tests
├── start_*.sh          # Auxiliary ROS 2 launch scripts
└── docs/               # Technical documentation

---

How to Run the Smart Mission

To start the autonomous workflow, the robot's logic must be executed on the robot itself, and the AI vision node on the Host PC.

1. On the Robot

Launch the unified mission script. This will start SLAM, the Nav2 stack, the video stream, and the UDP Listener.

./start_mission.sh

2. On the Host PC

Ensure the YOLO environment is set up properly. Run the vision node to start analyzing the warehouse environment:

python3 scripts/pc_vision_node.py

---

Utility Scripts

The repository includes several start_*.sh scripts in the root directory to facilitate ROS 2 lifecycle management.

• start_mission.sh: Main entry point. Starts SLAM localization, Nav2, Video streaming, Foxglove, and the UDP listener.
• start_slam_navigation.sh: Starts Navigation paired with SLAM Toolbox Localization.
• start_navigation.sh: Starts standard Navigation paired with AMCL.
• start_mapping.sh: Starts the SLAM Toolbox in mapping mode to create a new floor plan.
• save_map.sh: Utility to easily save a newly created map (e.g., ./save_map.sh my_new_map).
• send_goal.sh: Allows sending a specific navigation goal from the command line.
• kill_all_ros.sh: Emergency kill switch. Safely terminates all running ROS 2 nodes, video streams, and python scripts.

Manual Teleoperation

If manual control of the robot is required at any point, use the teleop node:

ros2 run teleop_twist_keyboard teleop_twist_keyboard

Saving a Map

If ./start_mapping.sh was used to create a new map, it can be saved easily by running the included utility script before shutting down the mapping nodes:

./save_map.sh my_new_map

---

Smart Vision Rules & Logistics

The Host PC continuously analyzes the camera feed and sends triggers to the Robot. The robot dynamically reacts to the following state changes:

• Pedestrian Zone: Drastically reduces linear speed and adjusts the pure pursuit lookahead distance to safely navigate around humans without oscillating.
• Restricted Area: Automatically skips the current waypoint and recalculates the route.
• Loading Zone: Halts the robot completely and waits for the user to send a continue signal to resume the mission.
• Stop for Safety: Performs an emergency halt for 5 seconds before resuming automatically.
• Robots-Only Zone: Restores navigation speed and lookahead distance to 100% nominal.
• Parking Zone: Halts the robot and safely initiates the shutdown sequence for the ROS 2 nodes.

---

Visualization and Monitoring

The system is pre-configured with modern monitoring tools to observe the robot's state from any computer on the network:

• Foxglove Studio: Enabled by default in the launch scripts. Connect to ws://<ROBOT_IP>:9090 using the Foxglove application to view live maps, topics, and telemetry.
• ROSboard: An alternative web-based visualizer. It is disabled by default to save resources, but can be enabled in the .sh launch files if needed.

---

Optional Docker Setup

If containerizing the robot's environment is desired, ensure the workspace is mounted properly.

# Example command to start and enter the container
docker start <your_container_name>
docker exec -it <your_container_name> bash

(Note: If deploying this code on our original Jetson Nano configuration, the container name is beautiful_snyder)