In [1]:
import numpy as np
import pprint
import sys
if "../" not in sys.path:
sys.path.append("../")
from lib.envs.gridworld import GridworldEnv
In [2]:
pp = pprint.PrettyPrinter(indent=2)
env = GridworldEnv()
In [6]:
# Taken from Policy Evaluation Exercise!
def policy_eval(policy, env, discount_factor=1.0, theta=0.00001):
"""
Evaluate a policy given an environment and a full description of the environment's dynamics.
Args:
policy: [S, A] shaped matrix representing the policy.
env: OpenAI env. env.P represents the transition probabilities of the environment.
env.P[s][a] is a (prob, next_state, reward, done) tuple.
theta: We stop evaluation one our value function change is less than theta for all states.
discount_factor: lambda discount factor.
Returns:
Vector of length env.nS representing the value function.
"""
# Start with a random (all 0) value function
V = np.zeros(env.nS)
while True:
delta = 0
# For each state, perform a "full backup"
for s in range(env.nS):
v = 0
# Look at the possible next actions
for a, action_prob in enumerate(policy[s]):
# For each action, look at the possible next states...
for prob, next_state, reward, done in env.P[s][a]:
# Calculate the expected value
v += action_prob * prob * (reward + discount_factor * V[next_state])
# How much our value function changed (across any states)
delta = max(delta, np.abs(v - V[s]))
V[s] = v
# Stop evaluating once our value function change is below a threshold
if delta < theta:
break
return np.array(V)
In [10]:
def policy_improvement(env, policy_eval_fn=policy_eval, discount_factor=1.0):
"""
Policy Improvement Algorithm. Iteratively evaluates and improves a policy
until an optimal policy is found.
Args:
env: The OpenAI envrionment.
policy_eval_fn: Policy Evaluation function that takes 3 arguments:
policy, env, discount_factor.
discount_factor: Lambda discount factor.
Returns:
A tuple (policy, V).
policy is the optimal policy, a matrix of shape [S, A] where each state s
contains a valid probability distribution over actions.
V is the value function for the optimal policy.
"""
# Start with a random policy
policy = np.ones([env.nS, env.nA]) / env.nA
policy_stable = True
while True:
V = policy_eval_fn(policy, env, discount_factor)
for s in range(env.nS):
old_action_distrib = policy[s]
action_values = []
# best action in this state (chosen greedily)
for a in range(env.nA):
q = 0
for prob, next_state, reward, done in env.P[s][a]:
q += prob * (reward + discount_factor * V[next_state])
action_values.append(q)
a = np.argmax(action_values)
policy[s] = np.array([1 if a == i else 0 for i in range(env.nA)])
if not np.array_equal(policy[s], old_action_distrib):
policy_stable = false
if policy_stable:
break
return policy, policy_eval_fn(policy, env, discount_factor)
In [11]:
policy, v = policy_improvement(env)
print("Policy Probability Distribution:")
print(policy)
print("")
print("Reshaped Grid Policy (0=up, 1=right, 2=down, 3=left):")
print(np.reshape(np.argmax(policy, axis=1), env.shape))
print("")
print("Value Function:")
print(v)
print("")
print("Reshaped Grid Value Function:")
print(v.reshape(env.shape))
print("")
In [12]:
# Test the value function
expected_v = np.array([ 0, -1, -2, -3, -1, -2, -3, -2, -2, -3, -2, -1, -3, -2, -1, 0])
np.testing.assert_array_almost_equal(v, expected_v, decimal=2)
In [ ]: