Skip to content

Latest commit

 

History

History
247 lines (176 loc) · 8.02 KB

advanced_installation.md

File metadata and controls

247 lines (176 loc) · 8.02 KB
Raw

Advanced installation and compilation of Black-DROPS code

We will go step by step on how to compile the Black-DROPS code and its dependencies by hand. Before proceeding it is recommended that you read the basic installation tutorial to get an idea of the process needed to be followed in general.

Black-DROPS is a limbo experiment

The Black-DROPS code is a limbo experiment (see the limbo docs for more details). This roughly means that the Black-DROPS code should be compiled within the exp folder of the limbo framework.

Installing limbo and its dependencies

Limbo has quite a few basic dependencies:

  • Eigen3
  • Boost
  • NLOpt (optionally but is recommended that we use it)

For Ubuntu-based distributions we should use the following commands to install Eigen3 and Boost:

sudo apt-get update
sudo apt-get install libeigen3-dev libboost-serialization-dev libboost-filesystem-dev libboost-test-dev libboost-program-options-dev libboost-thread-dev libboost-regex-dev

For OSX with brew:

brew install eigen
brew install boost

We highly recommend that you install NLOpt. Unfortunately, the Ubuntu packages do not provide NLOpt’s C++ bindings. You can get NLOpt here: http://ab-initio.mit.edu/wiki/index.php/NLopt [mirror: http://members.loria.fr/JBMouret/mirrors/nlopt-2.4.2.tar.gz].

For both Ubuntu and OSX systems, we should run something like the following:

cd /path/to/tmp/folder
wget http://members.loria.fr/JBMouret/mirrors/nlopt-2.4.2.tar.gz
tar -zxvf nlopt-2.4.2.tar.gz && cd nlopt-2.4.2
./configure -with-cxx --enable-shared --without-python --without-matlab --without-octave
sudo make install

If you want to install NLOpt somewhere else than /usr/local/, you add the --prefix=/path/to/install/folder in the ./configure command.

Now that the basic dependencies of limbo have been installed, we can proceed in cloning the limbo framework (we need the master branch):

cd /path/to/code/folder
git clone https://github.com/resibots/limbo.git
cd limbo
./waf configure
./waf

You can also clone with ssh if desired: git clone [email protected]:resibots/limbo.git.

Installing libcmaes

Next step is to install the libcmaes library. We have to first install the dependencies of the library.

For Ubuntu systems:

sudo apt-get install libgtest-dev autoconf automake libtool libgoogle-glog-dev libgflags-dev

For OSX:

brew install cmake autoconf automake

For Ubuntu systems, we need to make a minor fix for the gtest:

cd /usr/src/gtest
sudo mkdir -p build && cd build
sudo cmake ..
sudo make
sudo cp *.a /usr/lib

We also highly recommend to install TBB. For Ubuntu systems:

sudo apt-get install libtbb-dev

For OSX:

brew install tbb

We have slightly modified the library to suit our special use-case and thus we provide our own fork of the library:

cd /path/to/tmp/folder
git clone https://github.com/resibots/libcmaes.git
cd libcmaes
git checkout fix_flags_native

Now let's compile and install the library:

mkdir -p build && cd build
cmake -DUSE_TBB=ON -DUSE_OPENMP=OFF ..
make -j4
sudo make install

If you do not have TBB installed, you should use: cmake -DUSE_TBB=OFF -DUSE_OPENMP=ON ... If you want to install libcmaes somewhere else than /usr/local, you should use the -DCMAKE_INSTALL_PREFIX=/path/to/install/folder cmake argument.

Installing DART

If you want to compile and create DART-based scenarios, you will need to install the upstream DART by source.

For Ubuntu systems, please follow the detailed installation instructions on the DART documentation website. Make sure that you don't forget to add the PPAs as detailed here. What is more, you need to enable the -DART_ENABLE_SIMD flag in the CMake configuration. In addition, you need the following optional dependencies: DART Parsers and OpenSceneGraph GUI. Lastly, you need to checkout to the release-6.7 branch (and not the one provided in DART's documentation).

For Ubuntu <= 14.04 one more step is needed:

sudo apt-add-repository ppa:libccd-debs/ppa
sudo apt-add-repository ppa:fcl-debs/ppa

Here's what you should do:

sudo apt-add-repository ppa:dartsim/ppa
sudo apt-get update

sudo apt-get install build-essential cmake pkg-config git
sudo apt-get install libeigen3-dev libassimp-dev libccd-dev libfcl-dev libboost-regex-dev libboost-system-dev
sudo apt-get install libopenscenegraph-dev

sudo apt-get install libtinyxml-dev libtinyxml2-dev
sudo apt-get install liburdfdom-dev liburdfdom-headers-dev

cd /path/to/tmp/folder
git clone git://github.com/dartsim/dart.git
cd dart
git checkout release-6.7

mkdir build
cd build
cmake -DDART_ENABLE_SIMD=ON ..
make -j4
sudo make install

If you want to install DART somewhere else than /usr/local, you should use the -DCMAKE_INSTALL_PREFIX=/path/to/install/folder cmake argument.

For OSX systems with homebrew, please follow the detailed installation instructions on the DART documentation website. You need to follow the same procedure as for Ubuntu systems. In short you should do the following:

brew install eigen
brew install assimp
brew install homebrew/science/libccd
brew install dartsim/dart/fcl
brew install boost
brew install open-scene-graph

brew install tinyxml
brew install tinyxml2
brew install ros/deps/urdfdom

cd /path/to/tmp/folder
git clone git://github.com/dartsim/dart.git
cd dart
git checkout release-6.7

mkdir build
cd build
cmake -DDART_ENABLE_SIMD=ON ..
make -j4
sudo make install

The last step to be able to run and create DART-based scenarios with the Black-DROPS code is to install the robot_dart wrapper:

cd /path/to/tmp/folder
git clone https://github.com/resibots/robot_dart.git
cd robot_dart
git checkout osg_gui
./waf configure
./waf
sudo ./waf install

If you have installed DART in a different location than the default ones (i.e., other than /usr/local/ or /usr/), you the --dart=/path/to/dart/install option in ./waf configure.

Installing SDL2

For simple graphics, we use the SDL2 library.

For Ubuntu systems, you should:

sudo apt-get install libsdl2-dev libsdl2-image-dev

For OSX systems:

brew install sdl2

Installing simple_nn

For neural network policies, we are using the simple_nn library.

You should do the following:

cd /path/to/tmp/folder
git clone https://github.com/resibots/simple_nn.git
cd simple_nn
./waf configure
./waf
sudo ./waf install

Compiling the Black-DROPS code

We've now reached the point where we can compile the Black-DROPS code. You should do the following:

cd /path/to/limbo/folder
mkdir -p exp
cd exp
git clone https://github.com/resibots/blackdrops.git #(no need for recursive cloning)
cd ..
./waf configure --exp blackdrops
./waf --exp blackdrops -j4

Please remember that if you changed the installation location of DART, libcmaes or NLOpt, you should adapt your LD_LIBRARY_PATH.

And then every time you make a change to a source file (*.hpp or *.cpp), you should re-run the compilation command: ./waf --exp blackdrops -j4 (from the root of the limbo folder).

Running scenarios

  • All the executables including your own new scenarios (assuming the compilation produced no errors) should be located in the /path/to/limbo/build/exp/blackdrops/src/ folder
  • For example if we want to run the cartpole scenario without any visualization, we should use: ./path/to/limbo/build/exp/blackdrops/src/classic_control/cartpole_simu [args] (you can get help on what arguments to use, with /path/to/binary --help)