Impact of Computation in Integral Reinforcement Learning for Continuous-Time Control

This repository is the code of the paper "Impact of Computation in Integral Reinforcement Learning for Continuous-Time Control" submitted to ICLR 2024.

Environment Requirements

• Python
• Libraries:
  • numpy
  • matplotlib
  • scipy
  • probnum
  • sklearn
  • argparse
  • os

Examples

1. Simulations for trapezoidal rule for Example 1

   • File: example_tos_trapz.py

2. Simulations for Bayesian quadrature with Matérn kernel for Example 1

   • File: example_tos_Matern.py

3. Simulations for trapezoidal rule rule for Example 2

   • File: example_sin_trapz.py

4. Simulations for Bayesian quadrature with Matérn kernel for Example 2

   • File: example_sin_Matern.py

Subdirectories

• ThirdOrderSystem: Contains simulation results related to Example 1.
• SinSystem: Contains simulation results related to Example 2.
• Illustration_of_BQ: Contains scripts like BQ_matern.py and BQ_wiener.py that illustrate Bayesian quadrature methods.
• Motivation_Sinsystem: Motivated example for scenerios when the internal dynamics is unknown.
• Motivation_Cartpole: Motivated example for scenerios when the internal dynamics is known.