Skip to content

PhysicsGym: A Reinforcement learning interface for Partial differential equation control

License

Notifications You must be signed in to change notification settings

ajrox090/PhysicsGym

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PhysicsGym: A Reinforcement learning interface for Partial differential equation control

This repository contains the final submission code for Master's thesis: Developing a Reinforcement learning interface for partial differential equation control.

Usage

Define the following methods specific to each problem

1. Initialisation

  • obs_shape() : define the shape of your observation space
  • action_shape() : define the shape of your action space
  • physics : instantiate the physics object located in src/env/physics

2. Reset

  • build_obs() : calculate the observation vector in this method.
  • build_reward() : define reward calculation in this method.

3. Step

  • action_transform() : define the transformation according to the problem, e.g. if you want to applying actions only at certain parts of the domain.

Experiments

  • experiments/burgers_equation_experiments and experiments/heat_equation_experiments contains the PDE experiments for uncontrolled simulation, baseline agent, MPC agent and RL agent individually. The file ending in _eval compares all the three agents using multiple random initial states.

About

PhysicsGym: A Reinforcement learning interface for Partial differential equation control

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages