Skip to content

Latest commit

 

History

History
678 lines (560 loc) · 32.4 KB

Part3-Using-AirSim.md

File metadata and controls

678 lines (560 loc) · 32.4 KB

A minimalist guide to a few frameworks⁠—on Ubuntu and Vector computing⁠—for the person who was born yesterday

By a novice for the novice

Part 3 - Simulating Multiple UAVs with AirSim

The following instructions are for a fresh install of Ubuntu 18.04 LTS 64-bit on a Lenovo P52 (i7-8850H/Quadro P2000)

Everything after a $ is entered on a terminal; everything after >>> is passed to a Python interpreter

Refer to Appendix A for installation instructions

Using AirSim binaries on Ubuntu 18

AirSim (Documentation, GitHub) is an open source, photorealistic simulator for drones and ground vehicles developed on top of Epic's Unreal Engine 4 (UE4) by Microsoft Research (arXiv paper)

Concretely, AirSim is a UE4 Plugin—code and data you can add to any UE4 project; the simplest way to use AirSim on Ubuntu 18 is to:

  • Run the binaries of a precompiled UE4 environment containing the AirSim pluging
  • Control the robots through AirSim's Python APIs

Note 1: the steps to build the UE4 and UE4Editor (necessary to modify models and environments) and AirSim (including Python/C++ APIs and ROS wrapper nodes) from source code are given below

Note 2: binaries and sources can be run/built on Windows as well

Note 3: Epic Games Launcher can be used to download new maps/assets from UE4's Marketplace. It is not available on Linux (Windows and Mac only). Yet, assets can be transferred to Ubuntu once downloaded (see the next part of this guide) and 3rd party alternatives exist

Note 4: as of June 2020, support for Unity as an alternative to the UE4 is still experimental

Download the precompiled environments

Find the latest AirSim's Linux release—April 2020's v1.3.1 release includes the following environments:

Africa
Blocks
Building 99
Landscape Mountains
Neighborhood
Soccer Field
TrapCam
Zhangjiajie

Note: Windows' v1.3.1 precompiled binaries include different environments

Download all of environments (approx. 8GB)

$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/Africa.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/Blocks.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/Building_99.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/LandscapeMountains.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/Neighborhood.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/SoccerField.zip
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/TrapCam.zip.001
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/TrapCam.zip.002
$ wget https://github.com/microsoft/AirSim/releases/download/v1.3.1-linux/Zhangjiajie.zip

Note that TrapCam is split into two files that need to be concatenated into one before unzipping

$ cat TrapCam.zip.00* > TrapCam.zip

Run a precompiled environment

Use an environment, e.g. Africa.zip, by navigating to the download folder, unzipping, and running it

$ unzip Africa.zip -d Africa
$ cd Africa
$ ./Africa_001.sh              # each environment has a single ".sh" script in its top folder, the name can vary

Select "Yes" or "No" to start the environment with either a car or drone

Note: this prompt will disappear once you set SimMode in file ~/Documents/AirSim/settings.json (more on this below)

Once the simulation starts, use Alt+Enter to switch between full screen and window views

Note: if the environment's display window captures your cursor:

  • Use Alt+Tab to navigate back to the terminal and kill it with Ctrl+c
  • Or switch Ubuntu workspace (Ctrl+Alt+top or bottom arrow) and close it from the sidebar (right click and Quit)

Use Fn+F1 to display this list of commands

F1          Toggle this help
F3          Toggle wireframe whe you press F1
F10         Show weather options
F           Switch to FPV View
B           Switch to "fly with me"
\           Switch to ground observer view
/           Switch to chase with spring arm mode
M           Switch to manual camera control
                    Arrow keys
                    Page Up/Down -> move up, down
                    W, S -> pitch
                    A, D -> yaw
R           Toggle recording
;           Toggle debug report
0           Toggle all sub-windows
1           Toggle depth sub-window
2           Toggle segmentation sub-window
3           Toggle scene sub-window
T           Toggle trace line
Backspace   Reset everything

When starting an environment with a car (i.e., you answered "Yes" to the prompt), the arrows keys can be used to drive it

Use -ResX=num_x_pixels -ResY=num_y_pixels -windowed to set your preferred resolution

$ ./Africa_001.sh  -ResX=640 -ResY=480 -windowed

UE4.24 uses Vulkan drivers and ir can take up more GPU memory; if you get memory allocation errors, switch to OpenGL with

$ ./Africa_001.sh  -ResX=640 -ResY=480 -windowed -opengl

Note: the TrapCam environment is a quite complex one and it can take a few minutes to load: it simulates flora and fauna (see this paper) and it includes an interactive interface to adjust their configuration

Customize a simulation through settings.json

The first time you run AirSim, file ~/Documents/AirSim/settings.json will be created/initialized with

{
  "SeeDocsAt": "https://github.com/Microsoft/AirSim/blob/master/docs/settings.md",
  "SettingsVersion": 1.2
}

settings.json will be loaded everytime you load the AirSim plugin (i.e., running precompiled binaries or pressing "Play" in the UE4Editor)

settings.json can be used to configure many world's, vehicles', and cameras' parameters⁠—a complete list is provided here

For example, copy the following into settings.json to start a simulation

  • Using quadcopters ("SimMode": "Multirotor")
  • Containing 2 of them (named Drone0 and Drone1)
  • Setting them 2 meters apart along the X axis, and 1 along Y
  • And making the default ViewMode that of a GroundObserver
{
  "SeeDocsAt": "https://github.com/Microsoft/AirSim/blob/master/docs/settings.md",
  "SettingsVersion": 1.2,
  "SimMode": "Multirotor",
  "ViewMode": "GroundObserver",
  "Vehicles": {
    "Drone0": {
      "VehicleType": "SimpleFlight",
      "X": 0, "Y": 0, "Z": 0, "Yaw": 0
    },
    "Drone1": {
      "VehicleType": "SimpleFlight",
      "X": 2, "Y": 1, "Z": 0, "Yaw": 0
    }
  }
}

Note 1: VehicleType can be PhysXCar, SimpleFlight, PX4Multirotor, ComputerVision; there is no default: it must be specified

Note 2: SimpleFlight is AirSim's default flight controller (see its code here)

Note 3: "ViewMode": "noDisplay", can be used to freeze rendering while still exposing AirSim's APIs (including the ones for compute vision) and improving performance (i.e., frames per second)

Note 4: while a PhysicsEngineName setting exists, AirSim only supports PhysX for cars and FastPhysicsEngine for multirotors

Running the Africa environment with these settings.json will display 2 drones within it

$ ./Africa_001.sh  -ResX=640 -ResY=480 -windowed

Option A: use AirSim's Python APIs with Anaconda

AirSim exposes Python APIs to control vehicles—as of June 2020, Python 3.5 (or newer) and Anconda are recommended for their use

More thorough instructions to install and use coda are provided in here; these are minimal steps

$ sudo apt-get install libgl1-mesa-glx libegl1-mesa libxrandr2 libxrandr2 libxss1 libxcursor1 libxcomposite1 libasound2 libxi6 libxtst6
$ wget https://repo.anaconda.com/archive/Anaconda3-2020.02-Linux-x86_64.sh
$ bash Anaconda3-2020.02-Linux-x86_64.sh

Accept the terms and conditions; type yes; accept the default installation location by pressing Enter

When prompted to "Do you wish the installer to initialize Anaconda3 by running conda init?" type yes again

Finally, run

$ source ~/.bashrc
$ conda config --set auto_activate_base False          # this turns off the auto-activation of the (base) environment

Verify that you can exit and re-enter the default conda environment called (base)

$ conda deactivate
$ conda activate

Note: Ubuntu 18's default Pyhton is python3 (version 3.6) without pip; in conda's environment (base) the interpreter is called python (version 3.7) and pip is pre-installed⁠

In (base), install the inter-process messaging library MessagePack-RPC

$ conda activate
$ pip install msgpack-rpc-python

Note that this might throw an error and roll-back tornado from version 6.0.3 to 4.5.3—this is ok

ERROR: notebook 6.0.3 has requirement tornado>=5.0, but you'll have tornado 4.5.3 which is incompatible.
ERROR: distributed 2.11.0 has requirement tornado>=5; python_version < "3.8", but you'll have tornado 4.5.3 which is incompatible.
Installing collected packages: msgpack-python, tornado, msgpack-rpc-python
  Attempting uninstall: tornado
    Found existing installation: tornado 6.0.3
    Uninstalling tornado-6.0.3:
      Successfully uninstalled tornado-6.0.3
Successfully installed msgpack-python-0.5.6 msgpack-rpc-python-0.4.1 tornado-4.5.3

If tornado is accidentally upgraded later on, check this potential fix

Install AirSim's Python APIs (as of June 2020, v1.2.8) in (base) from the Python Package Index (PyPI)

$ conda activate
$ pip install airsim

Note: to modify AirSim's Python APIs, remember to eventually uninstall this package and use a local build (more on this below)

$ conda activate
$ pip uninstall airsim

AirSim's Python APIs 2-drone example

To run a 2-drone example, make sure that ~/Documents/AirSim/settings.json contains the configuration proposed above

Save the following 2 scripts in your home folder as ~/drone0.py and ~/drone1.py

import airsim                                                            # drone0.py

client = airsim.MultirotorClient()                                       # connect to the simulator
client.confirmConnection()
client.enableApiControl(True, vehicle_name="Drone0")                     # enable API control on Drone0
client.armDisarm(True, vehicle_name="Drone0")                            # arm Drone0

client.takeoffAsync(vehicle_name="Drone0").join()                        # let Drone0 take-off
client.moveToPositionAsync(15, -3, -4, 5, vehicle_name="Drone0").join()  # Drone0 moves to (15, -3, 4) at 5m/s and hovers (note the inverted Z axis)
client.hoverAsync(vehicle_name="Drone0").join()                          # .join() let the script wait for asynchronous (i.e. non-blocking) methods

raw = client.simGetImage("bottom_center", airsim.ImageType.Scene, vehicle_name="Drone0")           # take an image of type "Scene" from the "bottom_center" of "Drone0"
f = open("Drone0a.png", "wb")
f.write(raw)                                                             # save the image as a PNG file
f.close()

raw = client.simGetImage("bottom_center", airsim.ImageType.Segmentation, vehicle_name="Drone0")    # take an image of type "Segmentation" from the "bottom_center" of "Drone0"
f = open("Drone0b.png", "wb")
f.write(raw)                                                             # save the image as a PNG file
f.close()

airsim.wait_key('Press any key to reset')                                # press any key
client.armDisarm(False, vehicle_name="Drone0")                           # disarm Drone0
client.reset()                                                           # reset the simulation
client.enableApiControl(False, vehicle_name="Drone0")                    # disable API control of Drone0

And

import airsim                                                            # drone1.py

client = airsim.MultirotorClient()                                       # connect to the simulator
client.confirmConnection()
client.enableApiControl(True, vehicle_name="Drone1")                     # enable API control on Drone1
client.armDisarm(True, vehicle_name="Drone1")                            # arm Drone1

client.takeoffAsync(vehicle_name="Drone1").join()                        # let Drone1 take-off
client.moveToPositionAsync(20, 3, -1, 5, vehicle_name="Drone1").join()   # Drone1 moves to (20, 3, 1) at 5m/s and hovers (note the inverted Z axis)
client.hoverAsync(vehicle_name="Drone1").join()                          # .join() let the script wait for asynchronous (i.e. non-blocking) methods

raw = client.simGetImage("bottom_center", airsim.ImageType.Scene, vehicle_name="Drone1")      # take an image of type "Scene" from the "bottom_center" of "Drone1"
f = open("Drone1a.png", "wb")
f.write(raw)                                                             # save the image as a PNG file
f.close()

client.simSetCameraOrientation("bottom_center", airsim.to_quaternion(1.5, 0, 0), vehicle_name="Drone1")      # modify the orientation (in radians) for camera "bottom_center" of "Drone1"

raw = client.simGetImage("bottom_center", airsim.ImageType.Scene, vehicle_name="Drone1")      # take another image of type "Scene" from the "bottom_center" of "Drone1"
f = open("Drone1b.png", "wb")
f.write(raw)                                                             # save the image as a PNG file
f.close()

airsim.wait_key('Press any key to reset')                                # press any key
client.armDisarm(False, vehicle_name="Drone1")                           # disarm Drone1
client.reset()                                                           # reset the simulation
client.enableApiControl(False, vehicle_name="Drone1")                    # disable API control of Drone1

Open 3 terminals (3x Ctrl+Alt+t); in the first one, run an AirSim environment, e.g., for "Neighborhood"

$ cd Neighborhood
$ ./AirSimNH.sh  -ResX=640 -ResY=480 -windowed

In each of the other 2 terminals, run the Python scripts

$ conda activate
$ python ~/drone0.py

And

$ conda activate
$ python ~/drone1.py

The 2 drones will perform a short flight and create 4 PNG images in ~: Drone0a.png, Drone0b.png, Drone1a.png, Drone1b.png

Note 1: AirSim's APIs use North-East-Down (NED) coordinates (Z points up in UE4 but down in AirSim)

Note 2: latitude and longitude of (0,0,0) can be set in settings.json by adding

  "OriginGeopoint": {
    "Latitude": 47.641468,
    "Longitude": -122.140165,
    "Altitude": 122
  },

Note: if you add this at the end (i.e., before the last }) of settings.json, you must remove the trailing comma and add one at the end of the previous setting

Cameras APIs and settings

The scripts above use camera bottom_center; the available cameras are

Car and Multirotor
    front_center
    front_right
    front_left
    back_center
Car-only
    fpv
Multirotor-only
    bottom_center

The scripts above use 2 image types, airsim.ImageType.Scene and airsim.ImageType.Segmentation; the available image types are

  Scene = 0,                       # UE4 rendered view
  DepthPlanner = 1,                # ground truth, camera plan depth 
  DepthPerspective = 2,            # ground truth, projection ray depth 
  DepthVis = 3, 
  DisparityNormalized = 4,
  Segmentation = 5,                # ground truth using UE4's meshes
  SurfaceNormals = 6,
  Infrared = 7                     # currently, a mapping from object id to the 0-255 range

Script drone1.py uses client.simSetCameraOrientation(..) to modify the orientation of a camera on "Drone1"

The default resolution of a camera capture is 256x144 pixels, to change camera settings and resolution

  "CameraDefaults": {
      "CaptureSettings": [
        {
          "ImageType": 0,
          "Width": 1024,
          "Height": 576,
          "FOV_Degrees": 90,
          "AutoExposureSpeed": 100,
          "MotionBlurAmount": 0
        }
    ]
  },

Note 1: if you add this at the end (i.e., before the last }) of settings.json, you must remove the trailing comma and add one at the end of the previous setting

Note 2: "ImageType": 0, means that this will only affect the new resolution (1024x576) of airsim.ImageType.Scene

Note 3: more information about the computer vision API are available at these links: APIs, pfm

Note 4: the time of the day (affecting the sun's position) can be set through the APIs or settings.json

Note 5: the weather can be modified from the APIs (see the code)

Note 6: the clock speed can also be adjusted from settings.json

Before moving on to the next steps, remember to uninstall airsim from (base) to avoid conflicts with the local build of AirSim

$ conda activate
$ pip uninstall airsim

Building Unreal Engine 4 and AirSim from source

Building the Unreal Engine 4 (and Editor) and AirSim allows to use the latter's C++ APIs and their ROS wrapper node

The UE4Editor and AirSim source code are also necessary to create custom maps, robot models, flight controllers, new APIs, etc. as shown in the next part of this guide

Build UE4 and UE4Editor

The first step is to build the Unreal Engine 4.24 on Ubuntu 18, to do so, one needs

Follow these instructions to link your GitHub and Epic accounts

Once you have been added to Epic Games' organization as a developer, you will be able to git clone UE4's source code

As of June 2020, 4.25 is the latest release but we use 4.24, as recommended by AirSim's documentation

The following steps take a bit more than 1 hour on a Lenovo P52 (and throw a few warnings)

$ sudo apt install git
$ cd ~
$ git clone -b 4.24 https://github.com/EpicGames/UnrealEngine.git
$ cd UnrealEngine
$ ./Setup.sh                                                             # if prompted, say yes to register file types
$ ./GenerateProjectFiles.sh
$ make

Modify your ~/.bashrc file to add the location of UE4's binaries to the PATH environmental variable

$ echo  'export PATH=~/UnrealEngine/Engine/Binaries/Linux:$PATH' >> ~/.bashrc       # note the use of single quotes ' and the prefixing
$ source ~/.bashrc

You can now run $ UE4Editor from any folder/location

Note: UE4 versions >4.21 use Vulkan; on Ubuntu 16, extra steps might be required (if one does not want to use the deprecated OpenGL)

Build AirSim

The following steps build AirSim in just a few minutes (and throw plenty of warnings)

$ cd ~
$ git clone https://github.com/Microsoft/AirSim.git
$ cd AirSim
$ ./setup.sh    # insert your password, if asked; note that this will install clang-8 and python 2.7 among other Ubuntu packages
$ ./build.sh

Modify your ~/.bashrc file to add the location of AirSim's Python APIs to the PYTHONPATH environmental variable

$ echo  'export PYTHONPATH=~/AirSim/PythonClient:$PYTHONPATH' >> ~/.bashrc          # note the use of single quotes ' and the prefixing
$ source ~/.bashrc

You can now use >>> import airsim in a Python interpreter from any folder/location

First use of the UE4Editor, Blocks environment, and Python APIs

Run the Unreal Engine 4 Editor from any folder by simply typing

$ UE4Editor

The first time, this might take a few minutes; when the UE4Editor start, to use the default "Blocks" environment, follow these steps

  • Choose "Select or Create New Project"
  • Click on "More", then "Browse"
  • Navigate to ~/AirSim/Unreal/Environments/Blocks
  • Select Blocks.uproject and "Open"
  • When the "Convert Project" menu appears, select the "More options" button
  • Then, "Convert in-place"

Note: the next time you run UE4Editor, Blocks will appear among the recent projects in the "Select or Create New Project" menu

You can dismiss the "New plugin" notification and close the UE4Editor interface tour; some shaders will still be compiling

Important: go to "Edit -> Editor Preferences"; in the "Search" box, type "CPU" and ensure that the "Use Less CPU when in Background" is unchecked

Wait for shaders to finish compiling, then click the "Play" button (or Alt+p)

The settings from ~/Documents/AirSim/settings.json will be loaded (thus, you will see 2 drones if you followed the steps above)

Fn+F1 will show the in-game menu; open a terminal (Ctrl+Alt+t) to run one of the Python examples

$ conda activate
$ python ~/drone0.py

Press Esc to end the simulation

Option B: Compile an executable that uses AirSim's C++ APIs

The simplest way to use AirSim's C++ APIs is to duplicate one of the example projects (e.g. ~/AirSim/HelloDrone) and use the same CMake-based compiler tool chain

Copy ~/AirSim/HelloDrone into a new folder called NewDroneProject

$ cp -r ~/AirSim/HelloDrone ~/AirSim/NewDroneProject

Remove Visual Studio's files and replace the content of HelloDrone.cpp

$ rm ~/AirSim/NewDroneProject/HelloDrone.vcxproj ~/AirSim/NewDroneProject/HelloDrone.vcxproj.filters
$ gedit ~/AirSim/NewDroneProject/HelloDrone.cpp

With the following lines

#include <iostream>
#include "vehicles/multirotor/api/MultirotorRpcLibClient.hpp"

int main() {
    using namespace std;
    msr::airlib::MultirotorRpcLibClient client;

    cout << "Press Enter to enable API control" << endl; cin.get();
    client.enableApiControl(true);
    
    cout << "Press Enter to arm the drone" << endl; cin.get();
    client.armDisarm(true);

    cout << "Press Enter to takeoff" << endl; cin.get();
    client.takeoffAsync(5)->waitOnLastTask();

    cout << "Press Enter to move 5 meters in x direction with 1 m/s velocity" << endl; cin.get();  
    auto position = client.getMultirotorState().getPosition(); // from current location
    client.moveToPositionAsync(position.x() + 5, position.y(), position.z(), 1)->waitOnLastTask();

    cout << "Press Enter to land" << endl; cin.get();
    client.landAsync()->waitOnLastTask();
    return 0;
}

Duplicate HelloDrone's cmake folder into one called ~/AirSim/cmake/NewDroneProject

$ cp -r ~/AirSim/cmake/HelloDrone ~/AirSim/cmake/NewDroneProject

Modify the CMakeLists.txt file in this folder by replacing the occurrences of HelloDrone with NewDroneProject

$ sed -i 's/HelloDrone/NewDroneProject/g' ~/AirSim/cmake/NewDroneProject/CMakeLists.txt

Add folder NewDroneProject to the top level CMakeLists.txt file

$ echo  'add_subdirectory("NewDroneProject")' >> ~/AirSim/cmake/CMakeLists.txt

Finally, setup and build AirSim again

$ cd ~/AirSim
$ ./setup.sh
$ ./build.sh

The executable of NewDroneProject will be located in ~/AirSim/build_debug/output/bin/ and can be run as

$ ~/AirSim/build_debug/output/bin/NewDroneProject

Remember to first start the UE4Editor and Blocks—and press "Play"—for the NewDroneProject executable to connect

Option C: using AirSim with ROS Melodic

AirSim's also provides a ROS wrapper node around its C++ library called airsim_node

Install ROS Melodic

First, install ROS Melodic following these steps

$ sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'
$ sudo apt-key adv --keyserver 'hkp://keyserver.ubuntu.com:80' --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654
$ sudo apt update
$ sudo apt install ros-melodic-desktop-full
$ echo "source /opt/ros/melodic/setup.bash" >> ~/.bashrc
$ source ~/.bashrc
$ sudo apt install python-rosdep python-rosinstall python-rosinstall-generator python-wstool build-essential
$ sudo apt install python-rosdep
$ sudo rosdep init
$ rosdep update

Note: while this guide focusses on Ubuntu 18 and ROS Melodic, as of June 2020, AirSim still works with Ubuntu 16 and ROS Kinetic as well

Build AirSim ROS nodes

Install additional packages and dependencies

$ sudo apt install ros-melodic-mavros ros-melodic-mavros-extras ros-melodic-tf2-sensor-msgs        # tf2-sensor-msgs required as of Summer 2020
$ sudo apt-get install gcc-8 g++-8

Set gcc8 as the default gcc in Ubuntu 18

$ sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-7 700 --slave /usr/bin/g++ g++ /usr/bin/g++-7
$ sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-8 800 --slave /usr/bin/g++ g++ /usr/bin/g++-8

Finally, build AirSim's ROS nodes; add to ~/.bashrc and source the ~/AirSim/ros/ workspace

$ cd ~/AirSim/ros
$ catkin_make
$ echo  "source ~/AirSim/ros/devel/setup.bash" >> ~/.bashrc
$ source ~/.bashrc

Launch AirSim ROS nodes

Make sure that the content of ~/Documents/AirSim/settings.json includes

  • Line "SimMode": "Multirotor",
  • And at least one vehicle, e.g. Drone0
{
  "SeeDocsAt": "https://github.com/Microsoft/AirSim/blob/master/docs/settings.md",
  "SettingsVersion": 1.2,
  "SimMode": "Multirotor",
  "Vehicles": {
    "Drone0": {
      "VehicleType": "SimpleFlight",
      "X": 0, "Y": 0, "Z": 0, "Yaw": 0
    }
  }
}

If you do not do so, the output of $ roslaunch airsim_ros_pkgs airsim_node.launch will include

Exception raised by the API:
rpclib: function 'getMultirotorState' (called with 1 arg(s)) threw an exception. The exception contained this information: Vehicle API for 'ComputerVision' is not available. This could either because this is simulation-only API or this vehicle does not exist.

And the output of $ rostopic list lines such as

/airsim_node/ComputerVision/*
/airsim_node/PhysXCar/*
/airsim_node/SimpleFlight/*

Start and environment (e.g. $ ~/Africa/Africa_001.sh -ResX=640 -ResY=480 -windowed), then run one of AirSim's launch files

To use node airsim_node only

$ roslaunch airsim_ros_pkgs airsim_node.launch

To use both node airsim_node and node pid_position_node

$ roslaunch airsim_ros_pkgs airsim_all.launch

airsim_node and pid_position_node's source code mostly resides in 2 files: airsim_ros_wrapper.cpp and pd_position_controller_simple.cpp; if you modify them, remember to recompile with

$ cd ~/AirSim/ros
$ catkin_make

Note 1: airsim_node.launch includes static_transforms.launch which uses tf to publish NED to ENU coodinates

Note 2: airsim_all.launch includes airsim_node.launch, position_controller_simple.launch, and dynamic_constraints.launch (which sets constraints for maxiums speeds and the gimbal)

Note 3: rviz.launch starts ROS' 3D visualization tool rviz

Note 4: airsim_with_simple_PD_position_controller.launch does not work because it includes static_transforms.launch twice (this can be fixed by removing line 10, if necessary)

In a new terminal (Ctrl+Alt+t), check which nodes are running

$ rosnode list

In a new terminal (Ctrl+Alt+t), check which topics are being published and subscribed

$ rostopic list -v

Print Drone0's GPS coodinates topic

$ rostopic echo /airsim_node/Drone0/global_gps

In a new terminal (Ctrl+Alt+t), check which services are available

$ rosservice list

Let Drone0 take-off (use Tab to autocomplete the arguments after the service name takeoff)

$ rosservice call /airsim_node/Drone0/takeoff "waitOnLastTask: true"

Known bugs and limitations

AirSim's ROS nodes are still being developed; e.g., in pd_position_controller_simple.cpp

  1. On lines 334 and 338, replace 4 occurrences of vel_cmd_.twist.linear.z with vel_cmd_.twist.angular.z

  2. To get rid of [ERROR] [1234567890.1234567890]: [PIDPositionController] Waiting for odometry!, IF ~/Documents/AirSim/settings.json contains a single quadcopter named Drone0, replace line 64

airsim_odom_sub_ = nh_.subscribe("/airsim_node/odom_local_ned", 50, &PIDPositionController::airsim_odom_cb, this);

With line

airsim_odom_sub_ = nh_.subscribe("/airsim_node/Drone0/odom_local_ned", 50, &PIDPositionController::airsim_odom_cb, this);

Keep AirSim's source code up-to-date

AirSim is constantly being developed (e.g. this pull request for airsim_ros_pkgs is open as of June 2020)

To pull updated source code (i.e. overwriting your modification but not deleting newly created files) and re-compile it

$ cd ~/AirSim
$ git reset --hard origin/master
$ git pull origin master
$ ./setup.sh
$ ./build.sh
$ cd ./ros
$ catkin_make

Contributed by James Xu

List of AirSim's Python APIs

See Python_API.md

List of AirSim's C++ APIs

See CPP_API.md

List of AirSim's ROS topics and services

See ROS_Wrapper.md

Work carried out @ University of Toronto's Dynamic Systems Lab / Vector Institute / Mitacs Elevate