def __init__(
self,
worker_id,
env_path,
config_path,
reduced_actions=False,
docker_training=False,
):
super(AnimalAIWrapper, self).__init__()
self.config = ArenaConfig(config_path)
self.time_limit = self.config.arenas[0].t
self.env = UnityEnvironment(
file_name=env_path,
worker_id=worker_id,
seed=worker_id,
n_arenas=1,
arenas_configurations=self.config,
docker_training=docker_training,
)
lookup_func = lambda a: {"Learner": np.array([a], dtype=float)}
if reduced_actions:
lookup = itertools.product([0, 1], [0, 1, 2])
else:
lookup = itertools.product([0, 1, 2], repeat=2)
lookup = dict(enumerate(map(lookup_func, lookup)))
self.action_map = lambda a: lookup[a]
self.observation_space = gym.spaces.Box(0,
255, [84, 84, 3],
dtype=np.uint8)
self.action_space = gym.spaces.Discrete(len(lookup))
self.t = 0
)
config_dict = {
1: "1-Food.yaml",
2: "2-Preferences.yaml",
3: "3-Obstacles.yaml",
4: "4-Avoidance.yaml",
5: "5-SpatialReasoning.yaml",
6: "6-Generalization.yaml",
7: "7-InternalMemory.yaml"
}
env_file = config_dict[int(parser.parse_args().env)]
arena_config = ArenaConfig(
"/home/duju/animal_ai_olympics/duju_animal_ai_olympics/configs/" +
env_file)
exp_title = "Animal_AI_Baseline4_" + env_file.split(".")[0]
print(exp_title)
exp_dir = "/home/duju/animal_ai_olympics/duju_animal_ai_olympics/baseline_results/" + exp_title
os.makedirs(exp_dir)
txt_path = os.path.join(exp_dir, "results.txt")
action_dict = {
0: np.array([1, 0]), # forward
1: np.array([0, 2]), # left
2: np.array([0, 1]), # right
# 3 : np.array([2,0]) # backward
}