plstab 0.1.0

No project description provided

Plstab = plane-stabilization envirnoment

This is a simple env. based on OpenAI Gym library. Envirnoment for training reinforcement learning (RL) agents.

Example of code to launch enviroment

import plstab
import numpy as np
import gymnasium as gym

env = gym.make('PlaneStab-v0', render_mode='human', wind_mode=3) # create env
obs, info = env.reset(seed=42)

for _ in range(100):
   action = np.zeros(2) # sample action
   obs, reward, terminated, truncated, info = env.step(action) # step env


   if terminated or truncated: # reset if episode ends
      obs, info = env.reset()
     
   env.render()

env.close()

args:

name or arg	available value	default value
render_mode	"human", None	None
wind_mode	$ {0, 1, 2, 3} $	0
wind_mag	$x \in [0, \infty ]$	4.0
reward_fn	lambda X, Vx, Y, Vy, Phi, Omega: ...	None

wind mode provides to 1 of 4 wind models:

• 0: off (zero windspeed)
• 1: no_gust (wind without randonmness)
• 2: simple (simple gust model)
• 3: dryden (dryden's turbulence model)

Observation space:

8-dim numpy vector which contains: [X, Vx, Y, Vy, Phi, Omega, Wx, Wy], where

• X - x-component of pos
• Vx - x-component of velocity
• Y - y-component of pos
• Vy - y-component of velocity
• Phi - pitch angle of airplane
• Omega - derivative of Phi
• Wx - x-component of wind speed
• Wy - y-component of wind speed

Action space:

Actions are:

• rate of propeler [-1, 1] -> [0, max]
• angle of elevator [-1, 1] -> [-15, 15] (degrees)

Rewards:

By defalt:

-(Phi ** 2 + 0.1 * Omega ** 2 + 0.01 * alpha ** 2)