Docker Image and Utility for Onboard Computers

TL;DR: This repository contains a Docker image and a container interface for simulation and hardware experiments of drones.

The docker images provide a ready-to-use environment with pre-installed packages for drone simulation and hardware offboard experiments, for both x64 and arm64 platforms. We recommend using this image for both simulation on workstations and real-world experiments, as it greatly simplifies the setup process and ensures environment consistency across different machines. The environment details can be found in the Environment Details.

• Docker Image and Utility for Onboard Computers
  • Usage
    • Basic Usage
      • 国内用户
    • Use prebuilt packages
      • VINS-Fusion
      • Fast-LIO2
      • SITL with gazebo and PX4 (for x64 image only)
    • Connect to the container via SSH
    • Simulation with Custom Drone Models
  • Environment Details
    • Intel RealSense Depth Camera Support
    • LiDAR Support
    • OptiTrack Motion Capture Support
    • Neural Network Inference Support

Usage

Basic Usage

Setting everything up:

# clone the repository
git clone https://github.com/flyingbitac/fly_in_docker.git
# then pull the prebuilt image
python fly_in_docker/docker.py pull
# or download the resources and build the image locally (not recommended)
python fly_in_docker/docker.py build

Work with the container:

# start the container
python fly_in_docker/docker.py start -d /path/to/your/workspace # will be mounted at /root/ws/workspace
# enter the container
python fly_in_docker/docker.py enter
# stop the container
python fly_in_docker/docker.py stop

国内用户

对于国内用户，为了绕过 Docker Hub 的访问限制，可以使用阿里云的 ACR 服务。只需在拉取容器时添加 -a 或 --use_alibaba_acr 参数即可：

# 使用阿里云 ACR 服务拉取容器
python fly_in_docker/docker.py pull -a

此外，镜像构建的过程也需要从外部网站拉取许多依赖，因此建议使用代理来加速构建：

http_proxy=http://<proxy_address>:<proxy_port> python fly_in_docker/docker.py build
# or
https_proxy=http://<proxy_address>:<proxy_port> python fly_in_docker/docker.py build

Use prebuilt packages

VINS-Fusion

source ~/ws/vins/devel/setup.bash
roslaunch vins up.launch

Fast-LIO2

Before using Fast-LIO2, please modify the IP, subnet mask, and gateway of the HOST OS ethernet to connect to the Livox LiDAR:

• IP: 192.168.1.5
• Subnet Mask: 255.255.255.0
• Gateway: 192.168.1.1

Then run:

source ~/ws/livox_ws/devel/setup.bash
roslaunch fast_lio2 hardware_lidar_imu.launch ip:=192.168.1.1xx

SITL with gazebo and PX4 (for x64 image only)

source ~/ws/px4_setup.bash
export DISPLAY=:x # replace `:x` with your display number
roslaunch px4 mavros_posix_sitl.launch [gui:=false] # if you want gazebo to run in the headless mode

Connect to the container via SSH

ssh root@<ip address of the host machine> -p 2222 # default password is 'letmein'

Simulation with Custom Drone Models

To use custom drone models, you can specify the path to the model directory using the -c or --custom_model_path option when starting the container. The model directory should contain a valid PX4 drone model.

For example:

# start the container with custom drone models
python fly_in_docker/docker.py start \
  -d /path/to/your/workspace \
  -c /path/to/your/custom_model_1 \
  -c /path/to/your/custom_model_2

# enter the container
python fly_in_docker/docker.py enter

# then re-compile the PX4 firmware with the custom models
cd ~/ws/PX4-Autopilot && DONT_RUN=1 make px4_sitl_default gazebo

# setting up environment variables
source ~/ws/px4_setup.bash

# start the simulation with the custom model
roslaunch px4 mavros_posix_sitl.launch vehicle:=custom_model_1

Environment Details

• OS: Ubuntu 20.04
• ROS & MAVROS: Noetic
• PX4 Firmware: v1.15.4
• Python: 3.8.10
• Gazebo: 11 (with PX4 SITL support) (x64 image only)

Intel RealSense Depth Camera Support

• ros-noetic-cv-bridge
• ros-noetic-realsense2-camera
• librealsense2
• opencv-python
• pyrealsense2
• vins-fusion

LiDAR Support

• Livox-SDK2
• livox_ros_driver2
• Fast-LIO2
• livox_laser_simulation

OptiTrack Motion Capture Support

• vrpn packages prebuilt

Neural Network Inference Support

• pytorch 2.4.1-cpu
• torchvision 0.19.1-cpu
• torchaudio 2.4.1-cpu
• onnxruntime-cpu