Constrained Differential Dynamic Programming (CDDP) solver in C++

+---------------------------------------------------------+
|    ____ ____  ____  ____    _          ____             |
|   / ___|  _ \|  _ \|  _ \  (_)_ __    / ___| _     _    |
|  | |   | | | | | | | |_) | | | '_ \  | |   _| |_ _| |_  |
|  | |___| |_| | |_| |  __/  | | | | | | |__|_   _|_   _| |
|   \____|____/|____/|_|     |_|_| |_|  \____||_|   |_|   |
+---------------------------------------------------------+

This library is under active development. Star the repo ⭐ to stay updated on its progress and eventual release. I welcome any suggestions as I'm building this library to gain practical C++ experience.

Overview

This is an optimal control solver library using constrained differential dynamic programming (CDDP) written in C++. This library is particularly useful for mobile robot trajectory optimization and model predictive control (MPC).

The CDDP library solves problems in the form:

$$ \min_{\mathbf{U}} J(\mathbf{x}_0, \mathbf{U}) = \phi(\mathbf{x}_N) + \sum \ell(\mathbf{x}_k,\mathbf{u}_k) $$

$$ \mathrm{s.t.~} \mathbf{x}_{k+1} = \mathbf{f}(\mathbf{x}_k,\mathbf{u}_k) $$

$$ \quad \mathbf{g}(\mathbf{x}_k,\mathbf{u})_k\leq 0 $$

$$ \quad {\mathbf{x}}_{0} = \mathbf{x}{(t_0)} $$

$$ \quad k = 0,1,\cdots N-1 $$

Examples

Dubins Car

Simple car-like robot with velocity and steering control:

./examples/cddp_dubins_car // after building

Bicycle Model

Bicycle model with velocity and steering control:

./examples/cddp_bicycle // after building

Pendulum

Simple pendulum with torque control:

./examples/cddp_pendulum // after building

Cartpole

Cartpole with cart control:

./examples/cddp_cartpole // after building

Quadrotor

Quadrotor with thrust control:

./examples/cddp_quadrotor // after building

Manipulator

Manipulator with joint torque control:

./examples/cddp_manipulator // after building

Installation

Dependencies

• Eigen (Linear Algebra Library in CPP)

sudo apt-get install libeigen3-dev # For Ubuntu
brew install eigen # For macOS

Although the library automatically finds and installs the following dependencies via FetchContent, if you do not have ones, here is how you can install on your own.

• OSQP (QP solver) and osqp-cpp (C++ Wrapper for OSQP)

conda install -c conda-forge osqp # Optional

• libtorch : This library utilizes Torch for its underlying computations. It will be automatically installed during the build process.

• CUDA(Optional): If you want to leverage GPU acceleration for torch, ensure you have CUDA installed. You can download it from the NVIDIA website.

Building

git clone https://github.com/astomodynamics/cddp-cpp
cd cddp-cpp
mkdir build && cd build
cmake ..
make -j4
make test

ROS

If you want to use this library for ROS2 MPC node, please refer CDDP MPC Package. You do not need to install this library to use the package. MPC package will automatically install this library as a dependency.

References

• Zhaoming Xie, C. Karen Liu, and Kris Hauser, "Differential Dynamic Programming with Nonlinear Constraints," 2017 IEEE International Conference on Robotics and Automation (ICRA), 2017.
• Zhamming Xie's GitHub repository: https://github.com/ZhaomingXie/CDDP
• ZJU FAST Lab's GitHub repository: https://github.com/ZJU-FAST-Lab/SDQP
• Yuval Tassa's iLQG/DDP trajectory optimization: https://www.mathworks.com/matlabcentral/fileexchange/52069-ilqg-ddp-trajectory-optimization

Third Party Libraries

This project uses the following open-source libraries:

• OSQP (Apache License 2.0)
• osqp-cpp (Apache License 2.0)
• matplotlibcpp (MIT License)
• sdqp (MIT License)
• libtorch (BSD 3-Clause License)

Citing

If you use this work in an academic context, please cite this repository.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

TODO

• add python binding
• add state constraint handling
• add initial guess functionality based on path planner solution
• add automatic differentiation
• add nonlinear objective and constraint handling
• add parallelization
• add simulation and its plots
  • Quadrotor
  • Quadruped robot
  • Manipulator
  • Spacecraft
  • Humanoid