def main(open_plot=True):
# Setup MDP, Agents.
markov_game = RockPaperScissorsMDP()
ql_agent = QLearningAgent(actions=markov_game.get_actions())
fixed_action = random.choice(markov_game.get_actions())
fixed_agent = FixedPolicyAgent(policy=lambda s: fixed_action)
# Run experiment and make plot.
play_markov_game([ql_agent, fixed_agent], markov_game, instances=15, episodes=1, steps=40, open_plot=open_plot)
def choose_mdp(mdp_name, env_name="Asteroids-v0"):
'''
Args:
mdp_name (str): one of {gym, grid, chain, taxi, ...}
gym_env_name (str): gym environment name, like 'CartPole-v0'
Returns:
(MDP)
'''
# Other imports
from simple_rl.tasks import ChainMDP, GridWorldMDP, FourRoomMDP, TaxiOOMDP, RandomMDP, PrisonersDilemmaMDP, RockPaperScissorsMDP, GridGameMDP
# Taxi MDP.
agent = {"x":1, "y":1, "has_passenger":0}
passengers = [{"x":4, "y":3, "dest_x":2, "dest_y":2, "in_taxi":0}]
walls = []
if mdp_name == "gym":
# OpenAI Gym MDP.
try:
from simple_rl.tasks.gym.GymMDPClass import GymMDP
except:
raise ValueError("(simple_rl) Error: OpenAI gym not installed.")
return GymMDP(env_name, render=True)
else:
return {"grid":GridWorldMDP(5, 5, (1, 1), goal_locs=[(5, 3), (4,1)]),
"four_room":FourRoomMDP(),
"chain":ChainMDP(5),
"taxi":TaxiOOMDP(10, 10, slip_prob=0.0, agent=agent, walls=walls, passengers=passengers),
"random":RandomMDP(num_states=40, num_rand_trans=20),
"prison":PrisonersDilemmaMDP(),
"rps":RockPaperScissorsMDP(),
"grid_game":GridGameMDP(),
"multi":{0.5:RandomMDP(num_states=40, num_rand_trans=20), 0.5:RandomMDP(num_states=40, num_rand_trans=5)}}[mdp_name]
#!/usr/bin/env python
# Python imports.
import random
# Other imports.
import srl_example_setup
from simple_rl.agents import QLearnerAgent, FixedPolicyAgent
from simple_rl.tasks import RockPaperScissorsMDP
from simple_rl.run_experiments import play_markov_game
# Setup MDP, Agents.
markov_game = RockPaperScissorsMDP()
ql_agent = QLearnerAgent(actions=markov_game.get_actions())
fixed_action = random.choice(markov_game.get_actions())
fixed_agent = FixedPolicyAgent(policy=lambda s: fixed_action)
# Run experiment and make plot.
play_markov_game([ql_agent, fixed_agent],
markov_game,
instances=15,
episodes=1,
steps=40)