ROS 2 Humble Docker Environment for ADAS Applications on Jetson Orin Nano

This repository provides Dockerfiles and supporting resources to create containerized ROS 2 Humble environments specifically tailored for running ADAS (Advanced Driver Assistance Systems) applications on NVIDIA Jetson Orin Nano.

It includes containerized ROS 2 drivers for:

• Socket CAN
• Continental ARS404/408 Radar
• USB Cameras

Additionally, it provides a sample implementation of a containerized YOLO object detection model that performs real-time inference using a USB camera connected to the Orin Nano, currently optimized for CPU inference.

---

Getting Started

Dependencies (required on Jetson Orin Nano)

• NVIDIA Jetson Orin Nano: Tested on Jetson Orin Nano 8GB Developer Kit with 120 GB SSD. (Tanna Techbiz)
• Operating System: Ubuntu 22.04 (LTS) with NVIDIA Jetpack 6.2 (L4T 36.4.3)
• NVIDIA Drivers: Installed with Jetpack 6.2. Verify installation with nvidia-smi command.
• CUDA 12.6: Installed with Jetpack 6.2. Verify installation with nvcc --version command.
• ROS 2 Humble (Optional): Installed on the host system. Follow the ROS 2 installation guide for instructions.
• Docker: Ideally pre-installed with Jetpack 6.2. If Docker daemon errors occur, refer to Docker setup guide.
• CAN Device Drivers: Tested with Peak USB CAN based on SocketCAN (User Manual)
• USB Camera Drivers: Tested with Logitech C525 (USB Camera Setup Guide)

---

Docker Image

Prebuilt Docker Image

Pull the prebuilt Docker image from Docker Hub:

docker pull nikhil8490/ros2-humble-docker-orin

Build Docker Image

Alternatively, build your own Docker image from the Dockerfile:

git clone https://github.com/nikhilnair8490/ros2-orin-adas-docker.git
cd ros2-orin-adas-docker
docker image build -t <image-name> -f Dockerfile .

---

Setting Up Radar

Before starting the container, enable the CAN port and configure the radar:

• Connect USB camera and USB CAN device (connected to radar)

Enable CAN port (adjust can0 or can1 as necessary):

sudo ip link set can0 up type can bitrate 500000
sudo ip link set can0 up

Setup radar hardware (One-time configuration. Enables Objects detection with all extended properties):

cansend can0 200#F8000000089C0000

---

Running Containers

Once the Docker image is built, use the provided Docker Compose files to manage containers.

Radar and Camera Containers

Use docker-compose.jetson.sensors.yaml to run radar and camera sensor nodes:

# Start containers in the background
docker compose -f docker-compose.jetson.sensors.yaml up -d

# Enter sensors container shell
docker compose -f docker-compose.jetson.sensors.yaml exec ros_dev_radar bash

# Stop and remove containers
docker compose -f docker-compose.jetson.sensors.yaml down

YOLO Object Detection Container

Use docker-compose.jetson.ObjDetection.yaml to run YOLO object detection:

# Start YOLO container
docker compose -f docker-compose.jetson.ObjDetection.yaml up -d

# Enter container shell
docker compose -f docker-compose.jetson.ObjDetection.yaml exec ros_dev_yolo bash

# Stop and remove container
docker compose -f docker-compose.jetson.ObjDetection.yaml down

---

Repository Contents

Services

• ros_dev_radar: Launches radar sensor node (continental_ars408_socket_can.launch.xml). Reads from CAN interface, publishing detected objects.

  • Example Parameters:
    • interface: CAN device name (can0)
    • receiver_interval_sec: CAN message timeout interval
    • visualize_radar_markers: Enable RViz visualization topics

• ros_dev_cam: Launches USB camera node using usb_cam package with configurable resolution, FPS etc. (see config/params_1.yaml).

• ros_dev_yolo: Runs YOLOv11 object detection using CPU inference. Configurable settings include:

  • model: Path to YOLO .pt model file (also availabe in repo at jetson_ros_ws/src/yolo_ros/models)
  • input_image_topic: Source camera image topic
  • device: CPU or GPU (cpu or cuda:0) Currently only cpu is supported
  • threshold, iou: Detection confidence and overlap settings
  • imgsz_width, imgsz_height: Input image dimensions

---

ROS2 Topics

Camera Topics

• /camera1/camera_info
• /camera1/compressedDepth (depth camera required)
• /camera1/image_compressed
• /camera1/image_raw

CAN Topics

• /from_can_bus

Radar Topics

• /radar_objects_raw
• /radar_scan_raw
• /radar_state
• /radar_markers (RViz visualization)

YOLO Topics

• /yolo/dbg_image: Image with bounding boxes
• /yolo/detections: Object detection results

---

References

• ARS408 Driver
• USB Camera Driver
• YOLO ROS
• ROS2 V4L2 Camera
• ROS2 SocketCAN