⚠️ Disclaimer ⚠️

This is a work-in-progress library. As such, it only contains basic features. For any doubt, bug, problem or suggestion feel free to open an issue.

By now it has only been tested with Ubuntu 20.04 and Python 3.8.

This library contains tools to solve optimal control problems (OCPs) that deal with unmanned aerial manipulators (UAMs). It is strongly dependant on Crocoddyl, whose API is used to build the OCPs.

It has two principal pieces:

• Trajectory generator: It can be used to generate different maneuvers using any type of UAM. The OCP is easily specified by means of a YAML file (see the example section)

• Nonlinear model predictive controllers: It contains several nMPC controllers. They differ on the way the OCP inside the controller is built and how it is updated at every nMPC step.

It also contains an implementation of the Squash-box FDDP solver presented in this paper.

This is a C++ library. However, it can also be used within a Python environment since almost all classes have their corresponding Python bindings.

Crocoddyl has several dependencies. We need to install the following ones:

Follow the installation instructions. Install from robotpkg recommended.

Follow the installation instructions. Install from robotpkg recommended.

Use the forked version of the original repository. The devel branch of the forked version contains additional UAM models with different platform and robotic arm combination:

Platforms:

• Planar small hexacopter (370mm)
• Planar medium-sized hexacopter (680mm)
• Fully actuated hexacopter (Tilthex)

Robotic Arms:

• 2 DoFs
• 3 DoFs
• 5 DoFs (with spherical wrist)

Quick installation instructions:

cd <choose-your-path>
git clone https://github.com/PepMS/example-robot-data
cd example-robot-data
git submodule update --init
git checkout devel
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j6
sudo make install

Crocoddyl stands for Contact Robot Optimal Control by Differential Dynamic Library. It is a library to create and solve optimal control problems for robotics.

⚠️ Crocoddyl needs to be built from source. ⬇️⬇️Reason⬇️⬇️

EagleMPC has Crocoddyl as its main dependency (most of the classes contain Crocoddyl objects). Unfortunately, we cannot use the release packages given by the Crocoddyl team and we need to build it from source. This is due to the implementation of the Squash-box FDDP solver in the EagleMPC library. It inherits from the base class crocoddyl::SolverFDDP that has been slightly modified to consider different stopping criteria.

The Crocoddyl version you need to build is in this branch.

Overview of the different branches in the forked Crocoddyl repository:

• master : Even with its master parent branch
• devel : Even with its devel parent branch
• sbfddp: Even with the devel branch adding the modifications of the stopping criteria.

Quick installation instructions:

cd <choose-your-path>
git clone https://github.com/PepMS/crocoddyl.git
cd crocoddyl
git submodule update --init
git checkout sbfddp
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j6
sudo make install

Add the installation folder to your environment variables:

export PATH=/usr/local/bin:$PATH
export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig:$PKG_CONFIG_PATH
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
export PYTHONPATH=/usr/local/lib/python3/dist-packages:$PYTHONPATH

⚠️ If you want to reproduce the experiments of the paper Full-body torque-level Nonlinear Model Predictive Control, checkout the Crocoddyl repository to this tag.

YAML files are used to ease the creation of optimal control problems, to help in the definition of new multicopter platforms or to define the tunning parameters for the MPC controllers.

You can install this dependency by doing:

sudo apt-get install libyaml-cpp-dev

Clone this repo and build:

git clone https://github.com/PepMS/eagle-mpc.git
cd eagle-mpc
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j6

Install this library. By default it will be installed at usr/local/. If you want to install it somewhere else, the CMAKE_INSTALL_PREFIX from the CMakeLists.txtshould be modified accordingly. Then, to install do

sudo make install

⚠️ Problem with the system locale ⚠️ The Yaml parser uses the std::stod function to convert a string to a double. This function is locale dependant. Be sure to have set a locale that uses . as a decimal separator. To make sure of it you can run:

export LC_NUMERIC="en_US.UTF-8"

As this library contains Python bindings to its C++ code, we can run a python-based example.

python3 path/to/multicopter_mpc/examples/python/trajectory_generation.py