actor-critic-on-shell

• Simple RL implemented in Shell Script

• Maze as environment

• Actor Critic method

Setup

1. clone this repo

Usage

1. change params in config.txt, as you like

   • You can change maze in GRID

   In GRID,

   Symbols	Meaning
   o	Goal
   x	Fail
   #	Block
   others	normal

   as default value

2. run bash train.sh <MAX_EPISODE>

   • MAX_EPISODE is optional, should be given as integer

   After learning, you may get V & Q table, such as

   and the best move