def eval_genome(genome, config):
net = neat.nn.FeedForwardNetwork.create(genome, config)
fitnesses = []
for runs in range(runs_per_net):
sim = Simulator(testNum=runs)
# Run the given simulation for up to num_steps time steps.
while sim.shouldContinue():
inputs = sim.getInputActivations()
action = net.activate(inputs)
# Apply action to the simulated cart-pole
sim.update(action[0], action[1])
fitnesses.append(sim.getFitness())
# The genome's fitness is its worst performance across all runs.
return statistics.mean(fitnesses)
class SimulationRenderer(arcade.Window):
def __init__(self, isTeleop, neuralNet=None):
super().__init__(500, 500, title="Sim", resizable=False)
self.isTeleop = isTeleop
self.neuralNet = neuralNet
self.robotPos = 0
self.targetPos = 0
self.moveRight = 0
self.moveLeft = 0
self.simulator = Simulator(testNum=0)
arcade.set_background_color(arcade.color.WHITE)
self.setup()
arcade.run()
def on_draw(self):
""" Render the screen. """
arcade.start_render()
arcade.draw_text("Sim Time: " + str(round(self.simulator.simTime, 1)),
500,
500,
arcade.color.BLACK,
9,
width=130,
align="left",
anchor_x="right",
anchor_y="top",
font_name="Monaco")
arcade.draw_text(
"Right Motor: " +
str(round(self.simulator.rightMotor.getPercentOutput(), 2)) + "%",
500,
480,
arcade.color.BLACK,
9,
width=130,
align="left",
anchor_x="right",
anchor_y="top",
font_name="Monaco")
arcade.draw_text(
"Left Motor: " +
str(round(self.simulator.leftMotor.getPercentOutput(), 2)) + "%",
500,
460,
arcade.color.BLACK,
9,
width=130,
align="left",
anchor_x="right",
anchor_y="top",
font_name="Monaco")
#draw the robot
arcade.draw_rectangle_filled(
self.simulator.robotPosAbsolute[0] + 150,
self.simulator.robotPosAbsolute[1] + 150, #x,y,width,height
self.simulator.robotWheelBase * 2,
self.simulator.robotWheelBase * 2,
arcade.color.GRAY,
tilt_angle=math.degrees(self.simulator.robotPosAbsolute[2]))
arcade.draw_rectangle_filled(
self.simulator.robotPosAbsolute[0] + 150 +
10 * math.cos(self.simulator.robotPosAbsolute[2]),
self.simulator.robotPosAbsolute[1] + 150 + 10 *
math.sin(self.simulator.robotPosAbsolute[2]), #x,y,width,height
10,
10,
arcade.color.LIGHT_GRAY,
tilt_angle=math.degrees(self.simulator.robotPosAbsolute[2]))
#draw the target
arcade.draw_rectangle_filled(
self.simulator.targetPosAbsolute[0] + 150,
self.simulator.targetPosAbsolute[1] + 150, #x,y,width,height
self.simulator.robotWheelBase * 2,
self.simulator.robotWheelBase * 2,
arcade.color.GREEN,
tilt_angle=math.degrees(self.simulator.targetPosAbsolute[2]))
arcade.draw_rectangle_filled(
self.simulator.targetPosAbsolute[0] + 150 +
10 * math.cos(self.simulator.targetPosAbsolute[2]),
self.simulator.targetPosAbsolute[1] + 150 + 10 *
math.sin(self.simulator.targetPosAbsolute[2]), #x,y,width,height
10,
10,
arcade.color.LIGHT_GREEN,
tilt_angle=math.degrees(self.simulator.targetPosAbsolute[2]))
def on_key_press(self, key, modifiers):
"""
Called whenever a key is pressed.
"""
if (modifiers == arcade.key.MOD_CTRL or modifiers
== arcade.key.MOD_COMMAND) and (key == arcade.key.Q
or key == arcade.key.W):
sys.exit(
) # make command + Q work so the program can be quit more easily
if not self.isTeleop: #ignore keypresses if teleop is disabled
return
if key == arcade.key.A:
self.moveRight = 1
self.moveLeft = 0
elif key == arcade.key.S:
self.moveRight = 1
self.moveLeft = 1
elif key == arcade.key.D:
self.moveRight = 0
self.moveLeft = 1
def on_key_release(self, key, modifiers):
"""Called when the user releases a key. """
if not self.isTeleop: #ignore keypresses if teleop is disabled
return
if key == arcade.key.A or key == arcade.key.S or key == arcade.key.D:
self.moveRight = 0
self.moveLeft = 0
def setMovement(self, right, left):
self.moveRight = right
self.moveLeft = left
self.simulator.update(self.moveRight, self.moveLeft)
def getInputActivations(self):
return self.simulator.getInputActivations()
def setup(self):
# Set up your game here
pass
def update(self, delta_time):
if self.isTeleop: #update automatically when it is teleop mode
self.simulator.update(self.moveRight, self.moveLeft)
elif self.neuralNet != None:
cmd = self.neuralNet.activate(self.getInputActivations())
self.setMovement(cmd[0], cmd[1])
def sigmoid(self, x):
return 1 / (1 + math.exp(-x))
def stopRender(self):
arcade.close_window()
def startTeleop():
window = SimulationRenderer(True)
def startAI():
local_dir = os.path.dirname(__file__)
with open(local_dir + '/winner-net', 'rb') as f:
print(f)
winner_net = pickle.load(f)
simWindow = SimulationRenderer(False, winner_net)