PierreExeter

import gym
from gym import error, spaces, utils
from gym.utils import seeding

class TicTacEnv(gym.Env):
	metadata = {'render.modes': ['human']}


	def __init__(self):
		self.state = []

PierreExeter

import gym
import pybulletgym

env = gym.make('ReacherPyBulletEnv-v0')
env.render(mode="human")

for episode in range(20):
    state = env.reset()             
    rewards = []
    

PierreExeter

import gym
import pybullet_envs
import time

env = gym.make("CartPoleBulletEnv-v1")
env.reset()

while True:
    action = [1, 0]
    state, reward, done, info = env.step(action)

PierreExeter

import pybullet as p
import pybullet_data
import time

# start pybullet simulation
p.connect(p.GUI)

# reset the simulation to its original state
p.resetSimulation()

PierreExeter

import pybullet as p
import time
import pybullet_data

# Start pybullet simulation
p.connect(p.GUI)    
# p.connect(p.DIRECT) # don't render

# load urdf file path
p.setAdditionalSearchPath(pybullet_data.getDataPath()) 

PierreExeter

model = Sequential()
model.add(Dense(24, input_dim=n_states, activation='tanh'))
model.add(Dense(48, activation='tanh'))
model.add(Dense(n_actions, activation='linear'))
model.compile(loss='mse', optimizer=Adam(lr=alpha, decay=alpha_decay))

PierreExeter

# TRAINING PHASE
rewards = [] 

for episode in range(n_episodes):
    current_state = env.reset()
    current_state = discretize(current_state)

    alpha = get_alpha(episode)
    epsilon = get_epsilon(episode)

PierreExeter

# HYPERPARAMETERS
n_episodes = 1000           # Total train episodes
n_steps = 200               # Max steps per episode
min_alpha = 0.1             # learning rate
min_epsilon = 0.1           # exploration rate
gamma = 1                   # discount factor
ada_divisor = 25            # decay rate parameter for alpha and epsilon

# INITIALISE Q MATRIX
Q = np.zeros(buckets + (n_actions,)) 

PierreExeter

# define the number of buckets for each state value (x, x', theta, theta')
buckets = (1, 1, 6, 12)     

# define upper and lower bounds for each state value
upper_bounds = [
        env.observation_space.high[0], 
        0.5, 
        env.observation_space.high[2], 
        math.radians(50)
        ]

PierreExeter

import gym
import numpy as np
import math
import matplotlib.pyplot as plt

# CREATE ENVIRONMENT
env = gym.make('CartPole-v0')
n_actions = env.action_space.n
n_states = env.observation_space.shape[0]
print("Action space size: ", n_actions)