# Author: Marlos C. Machado
import utils
import random
import plotting
import numpy as np
from gridworld import GridWorld
if __name__ == "__main__":
# Read input arguments
args = utils.ArgsParser.read_input_args()
# Create environment
env = GridWorld(path=args.input)
num_states = env.get_num_states()
num_actions = len(env.get_action_set())
num_rows, num_cols = env.get_grid_dimensions()
# Sarsa(0):
gamma = 0.95
step_size = 0.1
num_steps_episode = []
for seed in range(args.num_seeds):
random.seed(seed)
num_steps_episode.append([])
q_values = np.zeros((num_states, num_actions))
for i in range(args.num_episodes):
s = env.get_current_state()
a = utils.epsilon_greedy(q_values[s])
num_steps = 0
while num_steps < args.max_length_ep and not env.is_terminal():
