Go to file
2022-07-25 09:51:08 +02:00
.gitignore initial commit 2022-07-22 11:49:30 +02:00
license add license ("BSD Zero Clause License") 2022-07-25 09:51:08 +02:00
main.py initial commit 2022-07-22 11:49:30 +02:00
model.py initial commit 2022-07-22 11:49:30 +02:00
readme.md initial commit 2022-07-22 11:49:30 +02:00
requirements.txt initial commit 2022-07-22 11:49:30 +02:00
run.py initial commit 2022-07-22 11:49:30 +02:00
train.py initial commit 2022-07-22 11:49:30 +02:00

Lunar Lander

This is an example of an Actor-Critic learning Agent as part of the ASIM RL Tutorial

It uses gym for the environment and torch as the basis for the A-C Network

Action Space

  1. do nothing,
  2. fire left engine,
  3. fire bottom engine,
  4. fire right engine

Observation Space

  1. x,
  2. y,
  3. linear velocity in x,
  4. linear velocity in y,
  5. angle,
  6. angular velocity
  7. ground contact leg 1
  8. ground contact leg 2

Rewards

  1. landing at landing pad +[100-140] points.
  2. crash: -100 points
  3. landing: +100 points
  4. landing on leg: +10 points per leg
  5. firing an engine: -0.3 points per engine per frame

it counts as solved at 200 points

Starting State

The lander starts at the top center of the viewport with a random initial force applied to its center of mass.