gui=True,
record=True)
elif ARGS.exp.split("-")[1] == 'payloadcoop':
test_env = PayloadCoop(
num_drones=NUM_DRONES,
aggregate_phy_steps=shared_constants.AGGR_PHY_STEPS,
obs=OBS,
act=ACT,
gui=True,
record=True)
else:
print("[ERROR] environment not yet implemented")
exit()
#### Show, record a video, and log the model's performance #
obs = test_env.reset()
logger = Logger(logging_freq_hz=int(test_env.SIM_FREQ /
test_env.AGGR_PHY_STEPS),
num_drones=NUM_DRONES)
if ACT in [
ActionType.ONE_D_RPM, ActionType.ONE_D_DYN, ActionType.ONE_D_PID
]:
action = {i: np.array([0]) for i in range(NUM_DRONES)}
elif ACT in [
ActionType.RPM, ActionType.DYN, ActionType.VEL, ActionType.XYZ_YAW
]:
action = {i: np.array([0, 0, 0, 0]) for i in range(NUM_DRONES)}
elif ACT in [ActionType.PID, ActionType.XY_YAW]:
action = {i: np.array([0, 0, 0]) for i in range(NUM_DRONES)}
elif ACT in [ActionType.JOYSTICK]:
action = {i: 0 for i in range(NUM_DRONES)}
record=False
)
elif ARGS.exp.split("-")[1] == 'meetup':
test_env = MeetupAviary(num_drones=NUM_DRONES,
aggregate_phy_steps=shared_constants.AGGR_PHY_STEPS,
obs=OBS,
act=ACT,
gui=True,
record=True
)
else:
print("[ERROR] environment not yet implemented")
exit()
#### Show, record a video, and log the model's performance #
obs = test_env.reset()
logger = Logger(logging_freq_hz=int(test_env.SIM_FREQ/test_env.AGGR_PHY_STEPS),
num_drones=NUM_DRONES
)
if ACT in [ActionType.ONE_D_RPM, ActionType.ONE_D_DYN, ActionType.ONE_D_PID]:
action = {i: np.array([0]) for i in range(NUM_DRONES)}
elif ACT in [ActionType.RPM, ActionType.DYN, ActionType.VEL]:
action = {i: np.array([0, 0, 0, 0]) for i in range(NUM_DRONES)}
elif ACT==ActionType.PID:
action = {i: np.array([0, 0, 0]) for i in range(NUM_DRONES)}
else:
print("[ERROR] unknown ActionType")
exit()
start = time.time()
for i in range(6*int(test_env.SIM_FREQ/test_env.AGGR_PHY_STEPS)): # Up to 6''
#### Deploy the policies ###################################
agent.restore(ARGS.file)
#### Extract and print policies ####################################################################
policy0 = agent.get_policy("pol0")
policy1 = agent.get_policy("pol1")
print(policy0.model.action_model)
print(policy0.model.value_model)
print(policy1.model.action_model)
print(policy1.model.value_model)
#### Create test environment ########################################################################
env = FlockAviary(num_drones=ARGS.num_drones,
gui=True,
record=True,
obstacles=True)
obs = env.reset()
action = {i: np.array([0, 0, 0, 0]) for i in range(ARGS.num_drones)}
start = time.time()
for i in range(10 * env.SIM_FREQ):
#### Deploy the policies ###########################################################################
# print("Debug Obs", obs)
temp = {}
temp[0] = policy0.compute_single_action(
np.hstack([obs[0], obs[1], action[1]]))
temp[1] = policy1.compute_single_action(
np.hstack([obs[1], obs[0], action[0]]))
# print("Debug Act", temp)
action = {0: temp[0][0], 1: temp[1][0]}
obs, reward, done, info = env.step(action)
env.render()