Resource-Gathering#
Action Space |
Discrete(4) |
Observation Shape |
(4,) |
Observation High |
[5 5 5 5] |
Observation Low |
[0 0 0 0] |
Reward Shape |
(3,) |
Reward High |
[0. 1. 1.] |
Reward Low |
[-1. 0. 0.] |
Import |
|
Description#
From “Barrett, Leon & Narayanan, Srini. (2008). Learning all optimal policies with multiple criteria. Proceedings of the 25th International Conference on Machine Learning. 41-47. 10.1145/1390156.1390162.”
Observation Space#
The observation is discrete and consists of 4 elements:
0: The x coordinate of the agent
1: The y coordinate of the agent
2: Flag indicating if the agent collected the gold
3: Flag indicating if the agent collected the diamond
Action Space#
The action is discrete and consists of 4 elements:
0: Move up
1: Move down
2: Move left
3: Move right
Reward Space#
The reward is 3-dimensional:
0: -1 if killed by an enemy, else 0
1: +1 if returned home with gold, else 0
2: +1 if returned home with diamond, else 0
Starting State#
The agent starts at the home position with no gold or diamond.
Episode Termination#
The episode terminates when the agent returns home, or when the agent is killed by an enemy.
Credits#
The home asset is from https://limezu.itch.io/serenevillagerevamped The gold, enemy and gem assets are from https://ninjikin.itch.io/treasure