updated stuff

environments/factory/factory_cleaning.py

@@ -0,0 +1,38 @@
+import numpy as np
+from pathlib import Path
+from environments import helpers as h
+
+
+class Factory(object):
+    LEVELS_DIR = 'levels'
+
+    def __init__(self, level='simple', n_agents=1, max_steps=1e3):
+        self.n_agents = n_agents
+        self.max_steps = max_steps
+        self.level = h.one_hot_level(
+            h.parse_level(Path(__file__).parent / self.LEVELS_DIR / f'{level}.txt')
+        )#[np.newaxis, ...]
+        self.reset()
+
+    def reset(self):
+        self.done = False
+        self.agents = np.zeros((self.n_agents, *self.level.shape))
+        free_cells = np.argwhere(self.level == 0)
+        np.random.shuffle(free_cells)
+        for i in range(self.n_agents):
+            r, c = free_cells[i]
+            self.agents[i, r, c] = 1
+        free_cells = free_cells[self.n_agents:]
+        self.state = np.concatenate((self.level[np.newaxis, ...], self.agents), 0)
+
+    def step(self, actions):
+        assert type(actions) in [int, list]
+        if type(actions) == int:
+            actions = [actions]
+        # level, agent 1,..., agent n,
+        for i, a in enumerate(actions):
+            h.check_agent_move(state=self.state, dim=i+1, action=a)
+
+if __name__ == '__main__':
+    factory = Factory(n_agents=1)
+    factory.step(0)

environments/factory/levels/simple.txt

@@ -0,0 +1,10 @@
+----------
+----------
+--------#-
+----------
+----------
+----------
+----#-----
+----------
+----------
+----------

environments/factory_cleaning.py

@@ -1 +0,0 @@
-import numpy as np

environments/helpers.py

@@ -0,0 +1,57 @@
+import numpy as np
+from pathlib import Path
+
+# Constants
+WALL = '#'
+
+
+# Utility functions
+def parse_level(path):
+    with path.open('r') as lvl:
+        level = list(map(lambda x: list(x.strip()), lvl.readlines()))
+    if len(set([len(line) for line in level])) > 1:
+        raise AssertionError('Every row of the level string must be of equal length.')
+    return level
+
+
+def one_hot_level(level, wall_char=WALL):
+    grid = np.array(level)
+    binary_grid = np.zeros(grid.shape)
+    binary_grid[grid == wall_char] = 1
+    return binary_grid
+
+
+def check_agent_move(state, dim, action):
+    agent_slice = state[dim]  # horizontal slice from state tensor
+    agent_pos = np.argwhere(agent_slice == 1)
+    if len(agent_pos) > 1:
+        raise AssertionError('Only one agent per slice is allowed.')
+    x, y = agent_pos[0]
+    x_new, y_new = x, y
+    if action == 0: # North
+        x_new -= 1
+    elif action == 1: # East
+        y_new += 1
+    elif action == 2: # South
+        x_new += 1
+    elif action == 3: # West
+        y_new -= 1
+    elif action == 4: # NE
+        x_new -= 1
+        y_new += 1
+    elif action == 5: # SE
+        x_new += 1
+        y_new += 1
+    elif action == 6: # SW
+        x_new += 1
+        y_new -= 1
+    elif action == 7: # NW
+        x_new -= 1
+        y_new -= 1
+
+
+
+if __name__ == '__main__':
+    x = parse_level(Path(__file__).parent / 'factory' / 'levels' / 'simple.txt')
+    y = one_hot_level(x)
+    print(np.argwhere(y == 0))