3D Detection AMR POC

This repository contains my ROS 2 AMR simulation and LiDAR-based 3D person detection proof of concept.

Project Goal

The goal is to move from an AGV-style robot concept toward an AMR by adding:

• Mapping
• Navigation
• 3D LiDAR perception
• AI person detection

System

The project uses:

• Ubuntu 22.04
• ROS 2 Humble
• Gazebo Classic
• Nav2
• RViz2
• OpenPCDet
• PointPillars

Repository Structure

3d_detection/

• amr_ws/
  • src/my_first_bot/
  • maps/
  • nav_config/
  • my_warehouse_map_new1.yaml
  • my_warehouse_map_new1.pgm
• .gitignore
• README.md

Current Status

Completed:

• Gazebo AMR simulation
• 3D LiDAR point cloud
• 2D map creation
• Nav2 navigation
• CVAT 3D labeling workflow
• PointPillars training
• Live person detection in RViz

AI Detection Workflow

Gazebo LiDAR point cloud -> ROS 2 bag recording -> PCD export -> CVAT 3D annotation -> OpenPCDet dataset conversion -> PointPillars training -> Live ROS 2 detection in RViz

Not Included

Large generated files are not uploaded to GitHub:

• build/
• install/
• log/
• ROS bags
• PCD datasets
• OpenPCDet checkpoints
• RTAB-Map .db files
• Python virtual environments

Basic Run

After cloning on a new PC:

1. Go to the workspace:

cd ~/github/3d_detection/amr_ws

2. Build:

colcon build

3. Source workspace:

source install/setup.bash

4. Run simulation:

ros2 launch my_first_bot full_sim.launch.py

Note

This project is a proof of concept, not a production safety system.