A framework for testing and benchmarking collision avoidance strategies.
CMake 3.1 or newer
Gazebo 8.0+ (for virtual camera/vehicle support)
socat 1.7+ (for testbed support)
GZSitl (for virtual vehicle support)
Collision Avoidance Library (Coav) is developed having drones in mind, so when compiling the library without additional options, features related to benchmark and simulation will be OFF by default. This should be the preferred way when you want to ship the library on your drone target/product.
Library Features and Options
Collision Avoidance Library has support to following features that can be defined on compile time:
|Feature/Option||Compile Options||Default Value|
|Intel RealSense support||WITH_REALSENSE||ON|
|Visual Debugger support||WITH_VDEBUG||OFF (depends on Gazebo)|
|Compile code samples||WITH_SAMPLES||OFF|Method 1 - Embed 'coav-control' on an Intel Aero image
This method is recommended for those who want to use 'coav-control' on an Intel Aero Drone.
A Yocto layer containing recipes to build and install coav-control can be found on the repository under the folder 'meta-coav'. This layer can be easily added to to the image build by following the steps described by Intel Aero documentation here. The recipe install the 'coav-control' utility tool as well an init script that runs the tool on start-up. You can change it's behavior by editing the script file at any point of the process that seems convenient to you (custom branch, custom recipe or changing the file on the drone itself).
自分でbuild & installを行う。
This method is recommended for those who will run simulations and tests on the development environment instead of a real drone. It is also recommend for those actively writing the library code because makes it easier to switch binaries for tests during development. If targeting an Intel Aero drone, check additional instructions about taking advantage of Yocto's SDK support.
If you want to use Gazebo, ensure you go through the instructions available here and ensure you install the libgazebo8-dev package;
sudo sh -c 'echo "deb http://packages.osrfoundation.org/gazebo/ubuntu-stablelsb_release -cs
main" > /etc/apt/sources.list.d/gazebo-stable.list'
deb http://packages.osrfoundation.org/gazebo/ubuntu-stable trusty main
sudo apt-get update
sudo apt-get install gazebo8
sudo apt-get install libgazebo8-dev
sudo apt-get instal libglm-dev python-future これだけでやって、後はskip
Install all build dependencies (the last two are needed to build librealsense):
sudo apt-get install git cmake libglm-dev python-future doxygen libusb-1.0-0-dev libglfw3-dev
Go through the steps to install librealsense which can be found here
The project use CMake as build system and does not support in-tree build. As such, create a separate folder before building.
Make sure you have initialized and updated all the required submodules at least once with:
git submodule update --init --recursive
Create a "build" folder and build the library using CMake as follows:
cmake .. -DWITH_GAZEBO=YES -DWITH_VDEBUG=YES -DWITH_TOOLS=YES
sudo make install
These instructions will build and install the targets on cmake's default install path (usually '/usr/local'). To modify the library options, the following syntax is used when issuing cmake:
cmake .. -D= -D=
Also, the following CMake options may be of value:
CMAKE_INSTALL_PREFIX Set a custom installation path. This path is also used for dependency search.
CMAKE_PREFIX_PATH Add paths to be searched when looking for dependencies
A more complete explanation of those options can be found on CMake's Documentation.
Search GLM and Mavlink on
Change the install path to
Compile the library additional tools (coav-sim)
cmake .. -DCMAKE_INSTALL_PREFIX= -DCMAKE_PREFIX_PATH= -DWITH_TOOLS=ON
Make sure that the library was compiled with 'Coav Tools' turned on. This will build a target coav-control that can be found in 'tools/coav-control/' inside the build folder.
coav-control can be used execute a simple collision avoidance system for a Mavlink controlled Quadcopter that is composed by: a sensor, a detection algorithm and a detection strategy. It needs interaction with an autopilot and a sensor to work, so it won't do much when executed alone.
The following will list the possible options for each component:
Obstacle detector based on 'Blob extraction'
'Stop' avoidance strategy
./coav-control -d DI_OBSTACLE -a QC_STOP -s ST_REALSENSE
For information on how to make use of 'Collision Avoidance Library' on simulated environment and how to take advantage of tests automation via testbed, please refer to the Simulation Docs.
Deploying on Intel Aero
Intel Aero firmware is based on Yocto, so the Yocto SDK for Intel Aero will be used to properly compile Collision Avoidance Library for deploy on Intel Aero.
Instruction on how to build Intel Aero image and the associated SDK can be found on Intel Aero Wiki. Intel Aero SDK will be missing one of the Collision Avoidance Library dependencies: GLM
Since GLM is a "headers only" library, cmake just need to know where to find the headers in order to successfully "cross-compile" it. This will be done with "-DCMAKE_PREFIX_PATH" parameter as described by the instructions bellow.
Once Intel Aero SDK is successfully installed, the following instructions will configure the environment and compile the library: source /environment-setup-core2-64-poky-linux mkdir build
cmake .. -DCMAKE_PREFIX_PATH=""
After a successful build, you can install Collision Avoidance Library in a temporary path:
make install DESTDIR=
tar cvf coav.tar *
Copy coav.tar to Intel Aero root dir and execute the following on Intel Aero:
[intel-aero]$ cd /
[intel-aero]$ tar xvf coav.tar
And Collision Avoidance Library should be successfully installed!