def agent():
env = RemoteEnv(IN_PORT=8080, OUT_PORT=7070)
env.open(0)
for i in range(10000000):
sum_energy = 0
state = env.step("restart")
prev_energy = state['energy']
print(prev_energy)
done = state['done']
while not done:
frame = image_decode(state['frame'], 20, 20)
print(frame)
action = int(input('action'))
for i in range(8):
state = env.step("act", action)
state = env.step('get_status')
done = state["done"]
energy = state['energy']
frame = state['frame']
touched = state['touched']
ID = state['id']
delta = (energy - prev_energy)
print(delta)
sum_energy += delta
prev_energy = energy
print(sum_energy)
env.close()
def step(self, action, value=None):
state = None
for i in range(8):
state = super().step(action, value)
if state['done']:
break
imgdata = image_decode(state['frame'], 20, 20)
self.buf.append(imgdata)
state['frame'] = np.array(self.buf, dtype=np.float32)
state['frame'] = np.moveaxis(state['frame'], 0, -1)
state['frame'] = state['frame'].reshape(1, 20, 20, 4)
self.proprioceptions[0] = state['touched']
self.proprioceptions[1] = state['energy']
state['proprioception'] = np.array(self.proprioceptions,
np.float32).reshape(
1, len(self.proprioceptions))
return state
def agent():
env = RemoteEnv(IN_PORT=8085, OUT_PORT=7075)
env.open(0)
for i in range(10000000):
state = env.step("restart")
prev_energy = state['energy']
done = state['done']
while not done:
action = int(input('action'))
#action = np.random.choice([0, 1])
reward_sum = 0
touched = -1
for i in range(8):
state = env.step("act", action)
state = env.step('get_status')
energy = state['energy']
if touched == -1:
touched = state['touched']
elif not state['touched'] in [0, 1]:
touched = state['touched']
if state['done']:
break
done = state['done']
reward_sum += (energy - prev_energy)
prev_energy = energy
frame = state['frame']
ID = state['id']
frame = image_decode(state['frame'], 20, 20)
print(frame)
print(reward_sum)
print(touched)
print(state['signal'])
print(state['done'])
prev_energy = energy
if i >= 2:
break
env.close()
def get_frame_from_fields(fields):
imgdata = image_decode(fields['frame'], 20, 20)
return imgdata
def step(self, action):
obs, reward, done, info = self.env.step(action)
return image_decode(obs), reward, done, info
def reset(self):
state = self.env.reset()
self.state = image_decode(state)