TIAGo Delivery Robot

An autonomous indoor delivery robot system for apartment buildings, built with Isaac Sim and ROS2.

Overview

Problem Statement

In delivery services, the "last mile" accounts for approximately 53% of total logistics costs. Within this segment, the "last 100 meters"—from a building's entrance to each residential unit—presents significant inefficiencies, particularly in Korean high-rise apartments.

Delivery drivers must navigate security entrances, wait for elevators, and traverse multiple floors repeatedly. A single apartment building delivery can occupy an elevator for 10-20 minutes, causing friction with residents. This creates problems for all parties: drivers waste time on building navigation, residents face security concerns with strangers in their building, and elevator congestion leads to conflicts.

Solution

We propose a dedicated indoor delivery robot that handles the last 100m of delivery:

1. Delivery drivers drop packages at a designated drop zone at the building entrance
2. The robot autonomously picks up packages, reads delivery labels via OCR, and navigates to each unit
3. Packages are placed at residents' doorsteps

This approach reduces driver workload, enhances building security, and eliminates elevator congestion issues.

Key Objectives

Objective	Description	Technology
Simulation Environment	Apartment environment with lobby, elevator, and corridors	Isaac Sim
Indoor Navigation	Autonomous navigation with obstacle avoidance and elevator handling	Nav2
Destination Recognition	Read apartment/unit numbers from delivery labels	OCR
Object Manipulation	Pick packages from drop zone and place at destinations	MoveIt
E2E Validation	Complete delivery scenario testing in simulation	Isaac Sim + ROS2

System Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        Orchestrator                              │
│                    (State Machine / Task Manager)                │
└──────────┬───────────────────┬───────────────────┬──────────────┘
           │                   │                   │
           ▼                   ▼                   ▼
    ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
    │  Navigation │     │ Perception  │     │Manipulation │
    │    (Nav2)   │     │ (OCR/Vision)│     │  (MoveIt)   │
    └─────────────┘     └─────────────┘     └─────────────┘
           │                   │                   │
           └───────────────────┴───────────────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Isaac Sim + ROS2  │
                    │   (TIAGo++ Robot)   │
                    └─────────────────────┘

Repository Structure

tiago-delivery/
├── README.md
├── dependencies.repos          # External dependencies (vcs)
├── setup.sh                    # Initial setup script
│
├── external/                   # External repos (gitignored)
│   └── tiago_isaac/            # TIAGo++ Isaac Sim simulation
│
├── ros2_ws/
│   └── src/
│       ├── tiago_rviz2/        # Symlink to tiago_isaac package
│       ├── simulation/         # Isaac Sim environments
│       ├── navigation/         # Nav2 configuration
│       ├── manipulation/       # MoveIt pick & place
│       ├── perception/         # Vision & OCR
│       ├── interfaces/         # Custom msg/srv/action
│       └── orchestrator/       # Task management
│
└── docker/                     # Docker configurations

Prerequisites

• AWS EC2 with NVIDIA GPU (g5.2xlarge or higher recommended)
• NVIDIA Isaac Sim 4.2.0
• Docker
• Python 3.10

Environment Setup (AWS EC2)

This project is designed to run on AWS EC2 with NVIDIA's official Isaac Sim AMI and Docker-based ROS2 Humble environment.

1. Launch EC2 Instance

1. AMI: Search for "NVIDIA Isaac Sim" in AWS Marketplace

   • Select the Ubuntu 24.04 based official NVIDIA AMI
   • This comes with Isaac Sim pre-installed at /opt/IsaacSim

2. Instance Type: g5.2xlarge or higher (RTX GPU required)

3. Storage: 200GB+ recommended

4. Security Group: Open the following ports

   Port	Protocol	Purpose
   22	TCP	SSH
   49100	TCP	Isaac Sim WebRTC Streaming
   8211	TCP	Isaac Sim Livestream
   8899	TCP	Isaac Sim Nucleus

2. Initial EC2 Setup

SSH into your instance and run:

# Update system
sudo apt update && sudo apt upgrade -y

# Install Docker (if not installed)
curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh
sudo usermod -aG docker $USER
newgrp docker

# Install vcstool (for dependency management)
pip install vcstool

3. Clone and Setup Project

cd /home/ubuntu
git clone https://github.com/C-2-Organization/tiago-delivery.git
cd tiago-delivery

# Fetch external dependencies (tiago_isaac)
./setup.sh

4. Build Docker Image

cd /home/ubuntu/tiago-delivery
docker build -t tiago-delivery-ros2 -f docker/Dockerfile.ros2-humble .

5. Configure Shell Aliases

Add aliases to your .bashrc for convenient usage:

echo "source /home/ubuntu/tiago-delivery/docker/aliases.sh" >> ~/.bashrc
source ~/.bashrc

This provides the following commands:

Alias	Description
isaac-sim	Launch Isaac Sim with streaming and TIAGo extension
tiago-ws	Start ROS2 Docker container
tiago-ws-build	Rebuild Docker image and start container

6. Build ROS2 Workspace

# Enter the container
tiago-ws

# Inside container: run setup and build
cd /home/ros/tiago-delivery
./setup.sh
cd ros2_ws
colcon build --symlink-install
source install/setup.bash

Running the System

Terminal 1: Launch Isaac Sim

# On EC2 host (not in Docker)
isaac-sim

Access the streaming viewer at:

https://<your-ec2-public-ip>:49100/streaming/webrtc-client

Then in Isaac Sim:

1. Open File → Open
2. Navigate to /home/ubuntu/tiago-delivery/external/tiago_isaac/
3. Load tiago_dual_functional_light.usd
4. Press Play to start simulation

Terminal 2: Launch ROS2 Nodes

# Enter container
tiago-ws

# Inside container
source /home/ros/tiago-delivery/ros2_ws/install/setup.bash

# Test robot control
ros2 topic pub /cmd_vel geometry_msgs/msg/Twist "{linear: {x: 0.5, y: 0.0, z: 0.0}, angular: {z: 0.0}}" --once

Verify Communication

# In container - check available topics
ros2 topic list

# Expected topics from Isaac Sim:
# /cmd_vel
# /joint_command
# /joint_states
# /tf
# /scan_front
# /scan_rear
# /gemini2/rgb
# /gemini2/depth

Installation (Local Development)

If you're running locally instead of EC2, ensure you have:

• Ubuntu 22.04
• ROS2 Humble
• NVIDIA Isaac Sim 5.0.0 installed
• NVIDIA RTX GPU

1. Clone the Repository

git clone https://github.com/C-2-Organization/tiago-delivery.git
cd tiago-delivery

2. Run Setup Script

The setup script will fetch external dependencies and configure symbolic links.

./setup.sh

This will:

• Clone tiago_isaac into external/ directory
• Create ros2_ws/src/ structure
• Set up symbolic links for TIAGo ROS2 packages

3. Build ROS2 Workspace

cd ros2_ws
colcon build --symlink-install
source install/setup.bash

4. Configure Isaac Sim Extension

Add the TIAGo extension path to Isaac Sim:

1. Open Isaac Sim
2. Navigate to Window → Extensions
3. Add extension search path:

   <project_root>/external/tiago_isaac/exts/
   

4. Enable omni.tiago_omnidirectional extension

Troubleshooting

Isaac Sim extension not loading

Ensure the extension path is correctly set:

# Check if extension folder exists
ls /home/ubuntu/tiago-delivery/external/tiago_isaac/exts/

# Manually add in Isaac Sim:
# Window → Extensions → Gear icon → Add search path

ROS2 topics not visible

# Verify ROS_DOMAIN_ID matches
echo $ROS_DOMAIN_ID  # Should be 110

# Check RMW implementation
echo $RMW_IMPLEMENTATION  # Should be rmw_cyclonedds_cpp

Robot not moving with /cmd_vel

1. Ensure Isaac Sim simulation is playing (not paused)
2. Verify omni.tiago_omnidirectional extension is enabled
3. Check the action graph is loaded in USD file

Robot Platform

This project uses the TIAGo++ robot simulation from AIS-Bonn/tiago_isaac.

Key features:

• Dual-arm mobile manipulator
• Mecanum wheel omnidirectional base
• RGB-D camera (Gemini2)
• Dual laser scanners
• Full ROS2 integration

Team

Team Name: TBD

Role	Member	Responsibilities
Simulation	-	Isaac Sim environment, robot setup
Navigation	-	Nav2 configuration, elevator integration
Manipulation	-	MoveIt pick & place
Perception	-	Box detection, grasp estimation, OCR
Integration	-	State machine, task orchestration

References

• TIAGo++ Isaac Sim
• Nav2 Documentation
• MoveIt 2 Documentation
• Isaac Sim Documentation

License

This project is licensed under the MIT License.

Note: The tiago_isaac dependency is licensed under AGPL-3.0.