def run_simulated_mission(model, display=None, use_delays=False):
print("Simulated mission running.")
world_model = WorldModel(BLUEPRINT, CONFIG_FILE, simulated=True)
ticks_left = 5 * MAX_EPISODE_TIME
total_reward = 0
current_r = 0
while (ticks_left > 0 and world_model.is_mission_running()):
ticks_left -= 1
current_r = world_model.reward()
action = model.act(current_r, world_model.get_observation())
if display is not None:
display.update(world_model)
total_reward += current_r
world_model.simulate(action)
if use_delays:
print(action)
time.sleep(ACTION_DELAY)
# Collect last reward, and give to model, then end the mission
current_r = world_model.reward()
model.act(current_r, world_model.get_observation())
total_reward += current_r
model.mission_ended()
print("Simulated mission ended")
return total_reward, (MAX_EPISODE_TIME - (ticks_left / 5))
def run_simulated_mission(model, mission, cfg, demo=False):
print("Simulated mission running.")
world_model = WorldModel(mission.blueprint,
cfg,
simulated=True,
agent_pos=mission.start_position)
ticks_left = 5 * mission.max_episode_time
total_reward = 0
current_r = 0
use_delays = mission.action_delay > 0
while (ticks_left > 0 and world_model.is_mission_running()):
ticks_left -= 1
current_r = world_model.reward()
if demo:
action = model.demo_act(world_model.get_observation())
else:
action = model.act(current_r, world_model.get_observation())
if mission.display is not None:
mission.display.update(world_model)
total_reward += current_r
world_model.simulate(action)
if use_delays:
print(action)
time.sleep(mission.action_delay)
# Collect last reward, and give to model, then end the mission
if mission.display is not None:
mission.display.update(world_model)
current_r = world_model.reward()
if not demo:
model.act(current_r, world_model.get_observation())
total_reward += current_r
model.mission_ended()
print("Simulated mission ended")
return MissionStats(reward=total_reward,
length=(mission.max_episode_time - (ticks_left / 5)))
