def part2b():
Agent = HW2Agent(env.build_map(), env.STATEMAP, 0.5, 0.95, 0.01)
Agent.value_iteration()
print("Agent B:")
print("\tNum evaluations: " + str(Agent.num_evals))
print(Agent.values)
x,y = Agent.generate_best_path()
plt.imshow(env.build_map())
plt.scatter(y,x)
plt.show()
plt.imshow(Agent.values)
plt.show()
def part2c():
Agent = HW2Agent(env.build_map(), env.STATEMAP, 0.02, 0.55, 0.01)
Agent.value_iteration()
print("Agent C:")
print("\tNum evaluations: " + str(Agent.num_evals))
print(Agent.values)
x,y = Agent.generate_best_path(stochastic=True, cutoff=1000)
plt.imshow(env.build_map())
plt.scatter(y,x)
plt.show()
plt.imshow(Agent.values)
plt.show()
def part1a():
Agent = HW2Agent(env.build_map(), env.STATEMAP, 0.02, 0.95, 0.01)
Agent.policy_iteration()
print("Agent A:")
print("\tNum improvements: " + str(Agent.num_improvements))
print("\tNum evaluations: " + str(Agent.num_evals))
print(Agent.values)
x,y = Agent.generate_best_path()
plt.imshow(env.build_map())
plt.scatter(y,x)
plt.show()
plt.imshow(Agent.values)
plt.show()
import matplotlib.pyplot as plt
from math import inf
from agent_class import HW3Agent
from build_environment import build_map, STATEMAP
MAP = build_map()
def test(episode_len, num_episodes, epsilon, default_r=-1):
Agent = HW3Agent(MAP, STATEMAP, epsilon, default_r=default_r)
Agent.on_policy_mc(episode_len, num_episodes)
y,x = Agent.generate_best_path(cutoff=1000, stochastic=True)
pathlen = len(x) if len(x) < 1000 else 'INF'
pathlen = "%d episodes with %d steps: %s" % (num_episodes, episode_len, str(pathlen))
print(pathlen)
plt.imshow(MAP)
plt.scatter(x,y)
title = "%d episodes of length %d" % (num_episodes, episode_len)
plt.suptitle(title, fontsize=15)
plt.title(pathlen)
#plt.show()
'''
for episodes in range(1,5):
for elen in range(1,5):
test(10 ** elen, 10 ** episodes, 0.02, default_r=-inf)
'''