def __init__(self, env, bot):
assert issubclass(env, NavEnvExt)
assert issubclass(bot, QLBot)
self.nav_env = env()
self.bot = bot(NFQ(), 5)
self.inter = Interaction(self.nav_env, self.bot)
self.inter_test = Interaction(self.nav_env, self.bot)
self.avg_rewards = []
self.reward_stds = []
self.test_performance()
class QLBotTest(object):
def __init__(self, env, bot):
assert issubclass(env, NavEnvExt)
assert issubclass(bot, QLBot)
self.nav_env = env()
self.bot = bot(NFQ(), 5)
self.inter = Interaction(self.nav_env, self.bot)
self.inter_test = Interaction(self.nav_env, self.bot)
self.avg_rewards = []
self.reward_stds = []
self.test_performance()
def test_performance(self):
avg, std = self.inter_test.compute_avg_reward(20, 50)
print("test the performance of the robot")
self.avg_rewards.append(avg)
self.reward_stds.append(std)
def run_one_cycle(self):
self.inter.env.__init__()
self.inter.interact_serie(500)
self.test_performance()
import sys
import os
sys.path.insert(0, os.path.abspath('..'))
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from inter.interaction import Interaction
from environment.nav_env_ext import NavEnvExt
from bot.explore_bot import ExploreBot
nav_env = NavEnvExt()
bot = ExploreBot()
inter = Interaction(nav_env, bot)
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(15, 5))
ax1.axis('off')
ax2.axis('off')
ax3.axis('off')
obs1 = nav_env.top_down_view(128)
obs2 = nav_env.egocentric_view(128)
obs3 = nav_env.show_observation(inter._observation)
im1 = ax1.imshow(obs1, origin='lower')
im2 = ax2.imshow(obs2, origin='lower')
im3 = ax3.imshow(obs3, interpolation='none', origin='lower')
def animate(*args):
inter.interact()
import sys
import os
sys.path.insert(0, os.path.abspath('..'))
import numpy as np
from inter.interaction import Interaction
from environment.nav_env_ego import NavEnvEgo
from bot.explore_bot import ExploreBot
from bot.state_repr_learn import StateReprLearn
from plot_exp_nav import plot_states
nav_env = NavEnvEgo()
bot = ExploreBot()
inter = Interaction(nav_env, bot)
inter.interact_serie(5000)
bot_srl = StateReprLearn(300, 2, bot.data)
# try to learn the representation
# bot_srl.gradient_descent(100)
# plot the result after n steps of gradient descent
def plot_x(n):
plot_states(np.array(inter.env_state_his),
bot_srl.states,
'x',
path="figures/nav_ego_{}_x".format(n))