def test_location_to_state():
"""Tests location_to_state maps each location to a unique integer"""
for num_rows in [12, 10]:
for num_cols in [15, 9]:
env = Four_Rooms_Environment(grid_width=num_cols, grid_height=num_rows)
observed_states = set()
for row in range(num_rows):
for col in range(num_cols):
state = env.location_to_state((row, col))
assert state not in observed_states
observed_states.add(state)
def test_check_user_location_and_goal_location_match_state_and_next_state():
"""Checks whether user location always matches state and next state correctly"""
for _ in range(50):
env = Four_Rooms_Environment()
env.reset()
for _ in range(50):
move = randint(0, 3)
env.step(move)
assert env.state == [env.location_to_state(env.current_user_location), env.location_to_state(env.current_goal_location)]
assert env.next_state == [env.location_to_state(env.current_user_location), env.location_to_state(env.current_goal_location)]
def test_location_to_state_and_state_to_location_match():
"""Test that location_to_state and state_to_location are inverses of each other"""
env = Four_Rooms_Environment(stochastic_actions_probability=0.0)
env.reset()
for row in range(env.grid_height):
for col in range(env.grid_width):
assert env.location_to_state((row, col)) == env.location_to_state(env.state_to_location(env.location_to_state((row, col))))
def test_lands_on_goal_correctly():
"""Checks whether getting to goal state produces the correct response"""
env = Four_Rooms_Environment(stochastic_actions_probability=0.0)
env.reset()
env.move_user(env.current_user_location, (3, 3))
env.move_goal(env.current_goal_location, (2, 2))
env.step(0)
assert env.reward == env.step_reward_for_not_achieving_goal
assert not env.done
env.step(3)
assert env.reward == env.reward_for_achieving_goal
assert env.done
env = Four_Rooms_Environment(stochastic_actions_probability=0.0)
env.reset()
env.move_user(env.current_user_location, (2, 3))
env.move_goal(env.current_goal_location, (2, 8))
for move in [2, 1, 1, 1, 1, 1, 0]:
env.step(move)
if move != 0:
assert env.reward == env.step_reward_for_not_achieving_goal
assert not env.done
else:
assert env.reward == env.reward_for_achieving_goal
assert env.done
def test_actions_execute_correctly():
"""Tests that actions execute correctly"""
env = Four_Rooms_Environment(stochastic_actions_probability=0.0)
env.reset()
env.move_user(env.current_user_location, (3, 3))
env.step(0)
assert env.current_user_location == (2, 3)
env.step(1)
assert env.current_user_location == (2, 4)
env.step(2)
assert env.current_user_location == (3, 4)
env.step(3)
assert env.current_user_location == (3, 3)
env.step(0)
assert env.current_user_location == (2, 3)
env.step(0)
assert env.current_user_location == (1, 3)
env.step(0)
assert env.current_user_location == (1, 3)
env.step(1)
assert env.current_user_location == (1, 4)
env.step(1)
assert env.current_user_location == (1, 5)
env.step(1)
assert env.current_user_location == (1, 5)
def test_randomness_of_moves():
"""Test that determine_which_action_will_actually_occur correctly implements stochastic_actions_probability"""
env = Four_Rooms_Environment(stochastic_actions_probability=0.0)
env.reset()
for _ in range(10):
for move in env.actions:
assert move == env.determine_which_action_will_actually_occur(move)
env = Four_Rooms_Environment(stochastic_actions_probability=1.0)
num_iterations = 10000
for move in env.actions:
moves = []
for _ in range(num_iterations):
moves.append(env.determine_which_action_will_actually_occur(move))
count = Counter(moves)
for move_test in env.actions:
if move != move_test: #We do this because stochastic probability 1.0 means the move will never be picked
assert abs((num_iterations / (len(env.actions)-1)) - count[move_test]) < num_iterations / 20.0, "{}".format(count)
env = Four_Rooms_Environment(stochastic_actions_probability=0.75)
num_iterations = 10000
for move in env.actions:
moves = []
for _ in range(num_iterations):
moves.append(env.determine_which_action_will_actually_occur(move))
count = Counter(moves)
for move_test in env.actions:
assert abs((num_iterations / len(env.actions)) - count[move_test]) < num_iterations / 20.0, "{}".format(count)
from A3C import A3C
from Agents.DQN_Agents.DQN_HER import DQN_HER
from DDQN import DDQN
from Environments.Four_Rooms_Environment import Four_Rooms_Environment
from Hierarchical_Agents.SNN_HRL import SNN_HRL
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DQN import DQN
config = Config()
config.seed = 1
config.environment = Four_Rooms_Environment(
15,
15,
stochastic_actions_probability=0.25,
random_start_user_place=True,
random_goal_place=False)
config.num_episodes_to_run = 1000
config.file_to_save_data_results = "Data_and_Graphs/Four_Rooms.pkl"
config.file_to_save_results_graph = "Data_and_Graphs/Four_Rooms.png"
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 3
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False