import random
import numpy
import sys
import time
import os
import json
from pong_environment_play import PongGame
policy_filename = "pong_policy.dat"
values_filename = "pong_values.dat"
alpha = 0.1 # values / critic learning parameter
beta = 0.1 # actor learning parameter
gamma = 0.9 # error signal: future states parameter
env = PongGame(["human", "computer"], bins=12)
world_dim = env.getWorldDim()
num_possible_moves = env.getActionDim()
state = env.getState(1)
pol_file = None
val_file = None
if os.path.exists(policy_filename):
pol_file = open(policy_filename, "r+")
policy = numpy.array(json.loads(pol_file.read()))
pol_file.close()
else:
# create random policy
import random from pong_environment_play import PongGame import time env = PongGame(["human", "computer"]) env.start() while True: time.sleep(0.05) direction = random.randint(-1, 1) outcome = 0 state, outcome = env.move(direction, 1)
# Add place cell to list
if DEBUG:
print("\tPlace cell mu:%f" % cell_mu)
cells.append(PlaceCell(cell_mu, cell_sigma, cell_min_rate, cell_max_rate, population_size, population_label_stem + ("_%u" % c), sdp_connection))
return cells
# ------------------------------------------
# PyNN entry point
# ------------------------------------------
# SpiNNaker setup
sim.setup(timestep=1.0,min_delay=1.0,max_delay=10.0)
# Create human vs ai pong game and start it
pong_game = PongGame(["human", "computer"])
pong_game.start()
# Create spinnaker pong player
player_thread = SpiNNakerPlayer(BOARD_ADDRESS, 0, pong_game, MAX_Y, UPDATE_TIMESTEP,
NUM_PADDLE_CELLS, PADDLE_CELL_SIGMA, CELL_MIN_RATE, PADDLE_CELL_MAX_RATE, PADDLE_CELL_POP_SIZE,
NUM_BALL_CELLS, BALL_CELL_SIGMA, CELL_MIN_RATE, BALL_CELL_MAX_RATE, BALL_CELL_POP_SIZE)
# Start player thread and simulation
player_thread.start()
sim.run(SIM_TIME)