def test_validActions2(self):
gameEnv2 = MAPF_Env.MAPFEnv(num_agents2,
world0=np.array(world2),
goals0=np.array(goals2),
DIAGONAL_MOVEMENT=False)
validActions2 = gameEnv2._listNextValidActions(1)
self.assertEqual(validActions2, [0])
# With diagonal actions
gameEnv2 = MAPF_Env.MAPFEnv(num_agents2,
world0=np.array(world2),
goals0=np.array(goals2),
DIAGONAL_MOVEMENT=True)
validActions2 = gameEnv2._listNextValidActions(1)
self.assertEqual(validActions2, [0, 5, 6, 7, 8])
def test_validActions1(self):
# MAPF_Env.MAPFEnv(self, num_agents=1, world0=None, goals0=None, DIAGONAL_MOVEMENT=False, SIZE=10, PROB=.2, FULL_HELP=False)
gameEnv1 = MAPF_Env.MAPFEnv(num_agents1,
world0=np.array(world1),
goals0=np.array(goals1),
DIAGONAL_MOVEMENT=False)
validActions1 = gameEnv1._listNextValidActions(1)
self.assertEqual(validActions1, [0, 1, 2])
# With diagonal actions
gameEnv1 = MAPF_Env.MAPFEnv(num_agents1,
world0=np.array(world1),
goals0=np.array(goals1),
DIAGONAL_MOVEMENT=True)
validActions1 = gameEnv1._listNextValidActions(1)
self.assertEqual(validActions1, [0, 1, 2, 5])
def test_validActions3(self):
gameEnv3 = MAPF_Env.MAPFEnv(num_agents3,
world0=np.array(world3),
goals0=np.array(goals3),
DIAGONAL_MOVEMENT=False)
validActions3a = gameEnv3._listNextValidActions(1)
validActions3b = gameEnv3._listNextValidActions(2)
self.assertEqual(validActions3a, [0])
self.assertEqual(validActions3b, [0, 2])
# With diagonal actions
gameEnv3 = MAPF_Env.MAPFEnv(num_agents3,
world0=np.array(world3),
goals0=np.array(goals3),
DIAGONAL_MOVEMENT=True)
validActions3a = gameEnv3._listNextValidActions(1)
validActions3b = gameEnv3._listNextValidActions(2)
self.assertEqual(validActions3a, [0, 5, 6, 7])
self.assertEqual(validActions3b, [0, 2, 5, 8])
def test_move_diag5(self):
#tests diag collisions
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=True)
s0 = gameEnv4.world.state.copy()
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 5))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.ACTION_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(1, gameEnv4.world.state[4, 4])
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(2, gameEnv4.world.state[3, 4])
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(2, gameEnv4.world.state[3, 4])
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 7))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.GOAL_REWARD)
self.assertFalse(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(1, gameEnv4.world.state[3, 3])
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(2, gameEnv4.world.state[3, 4])
#after waiting, we should be able to cross diagonally
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 0))
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.ACTION_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(2, gameEnv4.world.state[4, 3])
def test_validActions4(self):
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=False)
validActions4a = gameEnv4._listNextValidActions(1)
validActions4b = gameEnv4._listNextValidActions(2)
self.assertEqual(validActions4a, [0, 2])
self.assertEqual(validActions4b, [0, 2])
# With diagonal actions
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=True)
validActions4a = gameEnv4._listNextValidActions(1)
validActions4b = gameEnv4._listNextValidActions(2)
self.assertEqual(validActions4a, [0, 2, 5, 6, 7])
self.assertEqual(validActions4b, [0, 2, 5, 6])
def test_move_south2(self):
gameEnv2 = MAPF_Env.MAPFEnv(num_agents2,
world0=np.array(world2),
goals0=np.array(goals2))
s0 = gameEnv2.world.state.copy()
s1, r, d, _, o_g, _, _ = gameEnv2.step((1, 4))
s2 = gameEnv2.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertTrue(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def test_move_south1(self):
gameEnv1 = MAPF_Env.MAPFEnv(num_agents1,
world0=np.array(world1),
goals0=np.array(goals1))
s0 = gameEnv1.world.state.copy()
# return state, reward, done, nextActions, on_goal
s1, r, d, _, o_g, _, _ = gameEnv1.step((1, 4))
s2 = gameEnv1.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def testIdle1(self):
gameEnv1 = MAPF_Env.MAPFEnv(num_agents1,
world0=np.array(world1),
goals0=np.array(goals1))
s0 = gameEnv1.world.state.copy()
# return state, reward, done, nextActions, on_goal, blocking, valid_action
s1, r, d, _, o_g, _, _ = gameEnv1.step((1, 0))
s2 = gameEnv1.world.state.copy()
self.assertEqual(r, MAPF_Env.IDLE_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def test_move_south4b(self):
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4))
s0 = gameEnv4.world.state.copy()
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 4))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 4))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def test_move_west3b(self):
gameEnv3 = MAPF_Env.MAPFEnv(num_agents3,
world0=np.array(world3),
goals0=np.array(goals3))
s0 = gameEnv3.world.state.copy()
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv3.step((2, 3))
s2 = gameEnv3.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv3.step((1, 3))
s2 = gameEnv3.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def testIdle3(self):
gameEnv3 = MAPF_Env.MAPFEnv(num_agents3,
world0=np.array(world3),
goals0=np.array(goals3))
s0 = gameEnv3.world.state.copy()
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv3.step((1, 0))
s2 = gameEnv3.world.state.copy()
self.assertEqual(r, MAPF_Env.GOAL_REWARD)
self.assertFalse(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv3.step((2, 0))
s2 = gameEnv3.world.state.copy()
self.assertEqual(r, MAPF_Env.IDLE_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def test_move_southwest4a(self):
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=True)
s0 = gameEnv4.world.state.copy()
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 7))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.ACTION_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 7))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.COLLISION_REWARD)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def testIdle4(self):
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=False)
s0 = gameEnv4.world.state.copy()
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 0))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.GOAL_REWARD)
self.assertFalse(d)
self.assertTrue(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 0))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.IDLE_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
def test_move_diag2(self):
gameEnv4 = MAPF_Env.MAPFEnv(num_agents4,
world0=np.array(world4),
goals0=np.array(goals4),
DIAGONAL_MOVEMENT=True)
s0 = gameEnv4.world.state.copy()
# Agent 2
s1, r, d, _, o_g, _, _ = gameEnv4.step((2, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.ACTION_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(2, gameEnv4.world.state[4, 3])
# Agent 1
s1, r, d, _, o_g, _, _ = gameEnv4.step((1, 6))
s2 = gameEnv4.world.state.copy()
self.assertEqual(r, MAPF_Env.ACTION_COST)
self.assertFalse(d)
self.assertFalse(o_g)
self.assertEqual(np.sum(s0), np.sum(s2))
self.assertEqual(1, gameEnv4.world.state[4, 2])
lr = tf.constant(LR_Q)
trainer = tf.contrib.opt.NadamOptimizer(learning_rate=lr, use_locking=True)
if TRAINING:
num_workers = NUM_THREADS # Set workers # = # of available CPU threads
else:
num_workers = NUM_THREADS
NUM_META_AGENTS = 1
gameEnvs, workers, groupLocks = [], [], []
n = 1 #counter of total number of agents (for naming)
for ma in range(NUM_META_AGENTS):
num_agents = NUM_THREADS
gameEnv = mapf_gym.MAPFEnv(num_agents=num_agents,
DIAGONAL_MOVEMENT=DIAG_MVMT,
SIZE=ENVIRONMENT_SIZE,
observation_size=GRID_SIZE,
PROB=OBSTACLE_DENSITY,
FULL_HELP=FULL_HELP)
gameEnvs.append(gameEnv)
# Create groupLock
workerNames = ["worker_" + str(i) for i in range(n, n + num_workers)]
groupLock = GroupLock.GroupLock([workerNames, workerNames])
groupLocks.append(groupLock)
# Create worker classes
workersTmp = []
for i in range(ma * num_workers + 1, (ma + 1) * num_workers + 1):
workersTmp.append(Worker(gameEnv, ma, n, a_size, groupLock))
n += 1
workers.append(workersTmp)
