tanjunkai2001/ADP_lib

A C++ library supporting the implementation of various ADP algorithms

ADPlib - A C++ library supporting the implementation of various ADP algorithms for users and developers from different background.

This library contains numerical adaptive dynamic programming (ADP) algorithms to solve linear quadratic regulator (LQR) problems and algebraic Riccati equations (ARE). ADP aims at finding a stabilizing optimal control policy for dynamical systems via online learning. The aim of this project is to provide a complete implementation for online and offline ADP solvers that can be easily applied in the controller design for dynamical systems.

Learn more about LQR at: https://en.wikipedia.org/wiki/Linear–quadratic_regulator

Learn more about ARE at: https://en.wikipedia.org/wiki/Algebraic_Riccati_equation

Please visit my ResearchGate for more details about ADP: https://www.researchgate.net/profile/Tao_Bian2

Features

value iteration (VI) and policy iteration (PI) for continuous-time linear time invariant systems:

ALgorithmADP.h

VI and PI based ADP algorithms for continuous-time linear time invariant systems

ControllerADP.h
  ControllerVI.h
  ControllerPI.h

Different matrix classes and operations

Matrix.h
  SquareMatrix.h
    SymmetricMatrix.h
      Diagonal.h
MatrixCalc.h

Linear dynamical systems

Dynamical.h

{width=65%}

How to use:

A simple demo is provided in

main.cpp

Instruction for *nix systems:

make all
./ADPsolver

Future work:

1. discrete-time ADP
2. multi-thread implementation
3. nonlinear ADP