class WandererRoborobo(MultiAgentEnv):
action_space = Box(np.array([-1, -1], dtype=np.float32), np.array([1, 1], dtype=np.float32))
observation_space = Box(np.concatenate((np.repeat(0, 8), [-2, -15])), np.concatenate((np.repeat(1, 8), [2, 15])))
def __init__(self, nbrobs: int, max_moves: int):
super().__init__()
self.nbrobs = nbrobs
self.max_moves = max_moves
self.moves = 0
self.rob = None
self.wms = None
def __del__(self):
if self.rob:
self.rob.close()
def reset(self):
if self.rob is None:
self.rob = Pyroborobo('config/wanderer.properties', WorldObserver, "dummy", "dummy", "dummy",
{'gInitialNumberOfRobots': str(self.nbrobs)})
self.rob.start()
wo = self.rob.world_observer
wo.set_roborobo(self.rob)
self.wms = self.rob.world_models
else:
print('reset')
wo = self.rob.world_observer
wo.reset()
self.moves = 0
obs_dict = {'player'+str(i): self.get_obs(i) for i in range(self.nbrobs)}
return obs_dict
def get_obs(self, i):
return np.concatenate((self.wms[i].get_camera_sensors_dist(), [self.wms[i].translation, self.wms[i].rotation]))
def step(self, action_dict):
self.moves += 1
for i in range(self.nbrobs):
self.wms[i].translation = action_dict['player'+str(i)][0] * 2
self.wms[i].rotation = action_dict['player'+str(i)][1] * 15
stop = self.rob.update(1)
if stop:
sys.exit(0)
obs_dict = {'player'+str(i): self.get_obs(i) for i in range(self.nbrobs)}
punish = {i: (1 - obs_dict['player'+str(i)][2]) + (1 - obs_dict['player'+str(i)][3] + (1 - obs_dict['player'+str(i)][4])
+ (np.abs(obs_dict['player'+str(i)][9])/15))
for i in range(self.nbrobs)}
done = {'player'+str(i): self.moves > self.max_moves for i in range(self.nbrobs)}
done['__all__'] = all(done.values())
rewards = {'player'+str(i): self.wms[i].translation - 10*punish[i] for i in range(self.nbrobs)}
return obs_dict, rewards, done, {}
def render(self):
pass
class NegotiateRoborobo(MultiAgentEnv):
action_space = Box(np.array([-1] * 4, dtype=np.float32),
np.array([1] * 4, dtype=np.float32))
observation_space = Box(np.repeat(0, 8 * 4 + 2), np.repeat(1, 8 * 4 + 2))
def __init__(self, nbrobs: int):
super().__init__()
self.nbrobs = nbrobs
self.rob = None
self.wms = None
self.curlife = np.zeros(nbrobs, dtype=int)
def __del__(self):
if self.rob:
self.rob.close()
def reset(self):
if self.rob is None:
self.rob = Pyroborobo('config/pynegociate.properties', None, None,
None, None,
{'gInitialNumberOfRobots': str(self.nbrobs)})
self.rob.start()
wo = self.rob.world_observer
self.wms = self.rob.world_models
else:
print('reset')
wo = self.rob.world_observer
wo.reset()
obs_dict = {
f'{i}-{self.curlife[i]}': self.wms[i].get_observations()['obs']
for i in range(self.nbrobs)
}
return obs_dict
def step(self, action_dict):
for key in action_dict:
idplayer = _get_id_player(key)
self.wms[idplayer].set_actions(action_dict[key])
stop = self.rob.update(1)
if stop:
sys.exit(0)
obs_dict = {}
rewards_dict = {}
done_dict = {}
for i in range(self.nbrobs):
obs = self.wms[i].get_observations()
if obs['seeking']:
done = self.wms[i].get_done()
obs_dict[f'{i}-{self.curlife[i]}'] = obs['obs']
rewards_dict[f'{i}-{self.curlife[i]}'] = self.wms[
i].get_reward()
done_dict[f'{i}-{self.curlife[i]}'] = done
if done:
self.curlife[i] += 1
if done_dict:
done_dict['__all__'] = all(done_dict.values())
else:
done_dict['__all__'] = False
return obs_dict, rewards_dict, done_dict, {}
def render(self):
pass
