return random.choice(maximals)
def eval_one_step_prey(self, seeker, prey):
if seeker.caught(prey):
return prey
else:
choices = [
m
for m in get_neighbourhood(prey) if self.is_valid_position([m])
] + [prey]
return self.greedy_move_prey(choices) if choices else prey
if __name__ == "__main__":
rows = 60
cols = 80
g = GameBoard(rows, cols, (4, 4), (rows - 1, cols - 1))
episode_len = int((g.rows * g.cols) / 2)
episode_num = 50000
g.q_learning(0.7, 0.9, 0.5, episode_len, episode_num)
g.seeker.reset()
pos_list = g.eval_control(episode_len)
ani = Draw.AnimateGameBoard(g)
ani.show(pos_list)
for i in range(5):
seeker = g.seeker.get_random(g.rows, g.cols, g.is_valid_position)
if seeker: g.seeker = seeker
pos_list = g.eval_control(episode_len)
ani.show(pos_list)
if __name__ == "__main__":
rows = 30
cols = 40
rounds = 10 #cols*rows
episode_len = 100
episode_num = 5000
arena = seeker.GameBoard(rows, cols, (4, 4),
(int(rows / 2), int(cols / 2)))
trace_seeker = False
show = None
prey = shapes.Blob(arena.prey)
if trace_seeker:
Q = q_learning_seeker(arena, arena.seeker, arena.prey, 0.6, 0.9, 0.5,
episode_len, episode_num)
if arena.seeker.get_hash() not in Q: Q[arena.seeker.get_hash()] = 0
print("Q" + str(arena.seeker.get_pos()) + " := " +
str(Q[arena.seeker.get_hash()]))
movement = eval_seeker(arena, Q, arena.seeker, arena.prey, episode_len)
show = movement, [prey] * (len(movement))
else:
Q = q_learning_prey(arena, arena.seeker, arena.prey, 0.6, 0.9, 0.5,
episode_len, episode_num)
if prey.get_hash() not in Q: Q[prey.get_hash()] = 0
print("Q" + str(prey.get_pos()) + " := " + str(Q[prey.get_hash()]))
movement = eval_prey(arena, Q, arena.seeker, arena.prey, episode_len)
show = [arena.seeker] * len(movement), movement
ani = Draw.AnimateGameBoard(arena)
ani.show_exact(show)