def test_directional_sensors():
a = Animat()
lx, ly = 0.5, 0.5 # obj position
o = Obj(lx, ly, 'default')
a.add_obj(o)
res = np.linspace(0, 1, 50)
mesh = np.meshgrid(res, res)
def f(coords):
a.x = coords[0]
a.y = coords[1]
a.update_sensors()
return a.sensors['default'][Sides.LEFT]
zs = np.apply_along_axis(f, 0, mesh)
imshow(zs, extent=[0, 1, 0, 1], origin='lower')
plot(lx, ly, 'wo', label='object')
xlabel('animat position')
ylabel('animat position')
title('Animat is facing to the right')
legend()
show()
def test_animat_no_controller():
for n_animats in range(10):
duration = 50.0
DT = 0.02
iterations = int(np.round(duration / DT))
animat = Animat()
animat.x = np.random.randn()
animat.y = np.random.randn()
animat.a = np.random.rand() * np.pi * 2.0
obj = Obj(0, 0, 'default')
animat.add_obj(obj)
for iteration in range(iterations):
animat.calculate_derivative()
animat.euler_update(DT=DT)
left_sensor = animat.sensors['default'][Sides.LEFT]
right_sensor = animat.sensors['default'][Sides.RIGHT]
# print(f'l:{left_sensor}\t r:{right_sensor}')
# animat.lm = 0.4
# animat.rm = 0.5
animat.lm = left_sensor * 5
animat.rm = right_sensor * 5
plot(animat.x_h, animat.y_h, ',')
plot(animat.x_h[-1], animat.y_h[-1], 'ko', ms=3)
plot(-999, -999, 'k.', label='Aniamt Final Position')
plot(0, 0, ',', label='Animat Trajectory')
plot(0, 0, 'rx', label='Object Position')
xlim(-3, 3)
ylim(-3, 3)
legend()
gca().set_aspect('equal')
show()
def __init__(self, genome, parent=None, gen=0):
self.parent = parent
self.animat = Animat(genome)
self.gen = gen
# Mark whether the animat's phenotype needs updating.
self._dirty_phenotype = True
def simulate_trial(controller, trial_index, generating_animation=False):
# seed for environment for this generation
np.random.seed(generation_index * 10 + trial_index)
current_t = 0.0
score = 0.0
animat = Animat()
animat.x = 0.0
animat.y = 0.0
animat.a = 0.0
controller.trial_data = {}
foods = []
waters = []
traps = []
n = {ObjTypes.FOOD: 2, ObjTypes.WATER: 2, ObjTypes.TRAP: 2}
for obj_type in ObjTypes:
for _ in range(n[obj_type]):
x, y = random_obj_position(animat)
obj = Obj(x, y, obj_type)
animat.add_obj(obj)
if obj_type == ObjTypes.FOOD:
foods.append(obj)
if obj_type == ObjTypes.WATER:
waters.append(obj)
if obj_type == ObjTypes.TRAP:
traps.append(obj)
food_bat = 1.0
water_bat = 1.0
controller.trial_data['sample_times'] = []
controller.trial_data['water_battery_h'] = []
controller.trial_data['food_battery_h'] = []
controller.trial_data['score_h'] = []
controller.trial_data['eaten_FOOD_positions'] = []
controller.trial_data['eaten_WATER_positions'] = []
controller.trial_data['eaten_TRAP_positions'] = []
controller.trial_data['FOOD_positions'] = []
controller.trial_data['WATER_positions'] = []
controller.trial_data['TRAP_positions'] = []
for iteration in range(N_STEPS):
# battery states for plotting
controller.trial_data['sample_times'].append(current_t)
controller.trial_data['water_battery_h'].append(water_bat)
controller.trial_data['food_battery_h'].append(food_bat)
controller.trial_data['score_h'].append(score)
if generating_animation:
# used in animation
controller.trial_data['FOOD_positions'].append([(l.x, l.y)
for l in foods])
controller.trial_data['WATER_positions'].append([(l.x, l.y)
for l in waters])
controller.trial_data['TRAP_positions'].append([(l.x, l.y)
for l in traps])
current_t += DT
# drain battery states
DRAIN_RATE = 0.2
water_bat = water_bat - DT * DRAIN_RATE
food_bat = food_bat - DT * DRAIN_RATE
# if going faster drains more battery
# water_b -= (animat.lm**2) * DT * 0.01
# food_b -= (animat.rm**2) * DT * 0.01
score += (water_bat * food_bat) * DT
# pass sensor states to controller
for obj_type in ObjTypes:
controller.set_sensor_states(obj_type, animat.sensors[obj_type])
# set animat motor states with controller calculations
animat.lm, animat.rm = controller.get_motor_output(
(food_bat, water_bat))
animat.calculate_derivative() # calculate changes
animat.euler_update(DT=DT) # apply changes
# check for FOOD collisions
for food in animat.objs[ObjTypes.FOOD]:
if (animat.x - food.x)**2 + (animat.y - food.y)**2 < Obj.RADIUS**2:
food_bat += 20.0 * DT
controller.trial_data['eaten_FOOD_positions'].append(
(food.x, food.y))
food.x, food.y = random_obj_position(animat) # relocate entity
# check for WATER collisions
for water in animat.objs[ObjTypes.WATER]:
if (animat.x - water.x)**2 + (animat.y -
water.y)**2 < Obj.RADIUS**2:
water_bat += 20.0 * DT
controller.trial_data['eaten_WATER_positions'].append(
(water.x, water.y))
water.x, water.y = random_obj_position(
animat) # relocate entity
# check for TRAP collisions
for trap in animat.objs[ObjTypes.TRAP]:
if (animat.x - trap.x)**2 + (animat.y - trap.y)**2 < Obj.RADIUS**2:
food_bat -= 50.0 * DT
water_bat -= 50.0 * DT
score = 0.0
controller.trial_data['eaten_TRAP_positions'].append(
(trap.x, trap.y))
trap.x, trap.y = random_obj_position(animat) # relocate entity
# DEATH -- if either of the batteries reaches 0, the trial is over
if food_bat < 0.0 or water_bat < 0.0:
food_bat = water_bat = 0.0
break
# position of entities still not eaten at end of trial (used for plotting)
controller.trial_data['uneaten_FOOD_positions'] = [
(l.x, l.y) for l in animat.objs[ObjTypes.FOOD]
]
controller.trial_data['uneaten_WATER_positions'] = [
(l.x, l.y) for l in animat.objs[ObjTypes.WATER]
]
controller.trial_data['uneaten_TRAP_positions'] = [
(l.x, l.y) for l in animat.objs[ObjTypes.TRAP]
]
controller.trial_data['animat'] = animat
if TEST_GA:
# simple GA test -- maximise genome
score = np.mean(controller.genome)
return score
