A synthetic simple game. The agent moves along a 1D array. The agent starts at the middle of the array. It has two actions in each step: move left of right. Each move has a reward determined by MOVE_REWARD (-0.1). There are two rewards at both ends of the array. The left end has reward DEST_REWARD (4.0). The right end has reward DEST_REWARD/2 (2.0). The game ends if the agent arrives at either end. The optimal policy is to always move to the left.
- Python name:
simple_game
- C++ constructor name:
SimpleGame
- Name for the unified C++ simulator interface:
simple_game
Name | Description |
---|---|
array_size |
The array size. (Default: 0) |