Skip to content

Latest commit

 

History

History
43 lines (26 loc) · 1.45 KB

README.md

File metadata and controls

43 lines (26 loc) · 1.45 KB

Ex3EKF gymnasium environment

A gymnasium environment for a noisy master-slave system. This environment can be used to train an RL-based stationary Kalman filter.

Observation space

  • hat_x_1: The estimated angle.
  • hat_x_2: The estimated frequency.
  • x_1: Actual angle.
  • x_2: Actual frequency.

Action space

  • u1: First action coming from the RL Kalman filter.
  • u2: Second action coming from the RL Kalman filter.

Episode termination

An episode is terminated when the maximum step limit is reached, or the step cost exceeds 100.

Environment goal

The agent's goal in the Ex3EKF environment is to act so that the estimator estimates the original noisy system perfectly. By doing this, it serves as an RL-based stationary Kalman filter.

Cost function

The Ex3EKF environment uses the following cost function:

$$ cost = (hat_x_1 - x_1)^2 + (hat_x_2 - x_2)^2 $$

Environment step return

In addition to the observations, the environment returns an info dictionary containing the current reference and the error when a step is taken. This results in returning the following array:

[hat_x_1, hat_x_2, x_1, x_2, info_dict]

How to use

This environment is part of the Stable Gym package. It is therefore registered as a gymnasium environment when you import the Stable Gym package. If you want to use the environment in stand-alone mode, you can register it yourself.