EmbersArc/SCpp: Implementation of SCvx, linear MPC and LQR for trajectory generation and tracking · GitHub

This library implements various optimal control algorithms that are particularly suited for aerospace applications.

• Efficient Successive Convexification, a real-time guidance algorithm for optimal trajectory planning of constrained dynamical systems
• Generic linear receding-horizon SOCP MPC algorithm
• Linear Quadratic Regulator

• JIT derivative code generation with CppAD/CppADCodegen
• Intuitive interface to implement custom models
• Rapid iteration with parameters files

• 2D Rocket Model
• Rocket Landing Model with Quaternion
• Rocket Landing Model with Euler Angles

• C++17
• Eigen
• Boost (odeint and ptree)
• fmt (included as submodule)
• CppAD/CppADCodegen (included as submodule)
• Epigraph (included as submodule)

git clone --recurse-submodules https://github.com/EmbersArc/SCpp.git
cd SCpp
mkdir build
cd build
cmake ..
make

Available executables are:

• LQR_sim to simulate a trajectory with the classic MPC controller

• MPC_sim to simulate a trajectory with the classic MPC controller

• SC_oneshot to calculate one trajectory with Successive Convexification

• SC_sim to simulate a trajectory with Successive Convexification

Calculated trajectories are written to the output/<modelname> directory.

See existing models in the socp_models folder for some examples.

• Successive Convexification: A Superlinearly Convergent Algorithm for Non-convex Optimal Control Problems

• Successive Convexification for 6-DoF Mars Rocket Powered Landing with Free-Final-Time

(click on videos for higher quality versions)

feed-forward input tested in a box2d physics simulation

I'm looking forward to contributions, both problem formulations and improvements to the core library.