def update(self, dt):
self.update_data()
self.data.pose = utils.update_pose(self.data.pose, self.data.wheel_speeds, dt)
self.data.x_positions.append(self.data.pose.x)
self.data.y_positions.append(self.data.pose.y)
self.data.theta = self.data.pose.theta
self.particle_filter.update(dt)
self.ppose = self.particle_filter.most_likely()
if self.plotting:
self.fig.canvas.draw()
def update(self, dt):
t0 = time.time()
particles = self.sample_particles(len(self.particles))
#Move sampled particles based on a motor control
new_particles = [
utils.update_pose(p, self.data.wheel_speeds, dt, noisy=True)
for p in particles
if p.x < 1469 and p.x > 0
and p.y < 962 and p.y > 0]
# odo_pose = self.data.pose
# odo_pose.w = 1./len(new_particles)*10
# new_particles.append(odo_pose)
if len(new_particles) < self.n:
shortage = self.n - len(new_particles)
likliest_particle = self.most_likely()
new_particles += (self.rand_gaussian_particles(likliest_particle,
shortage, likliest_particle.w))
eta = 0
#Calculate the sensor probabilities (weights) for each particle
#Use a pool of workers to utilize multiple cores
exp_readings = self.pool.map(raycasting.exp_readings_for_pose_star, itertools.izip(new_particles, itertools.repeat(self.data.distance_thresholds)))
if len(exp_readings) != len(new_particles):
assert False, "Array of expected readings must have the same size as the array of new particles. exp_readings: %d new_particles: %d" % (len(exp_readings), len(new_particles))
#Sum sensor probabilites (assumption is that they are independent)
for i, p in enumerate(new_particles):
weight = self.probability_sum(4, exp_readings[i])
eta += weight
p.w = weight
#Normalize weights
print 'eta: ', eta
new_particles = self.normalize(eta, new_particles)
while len(new_particles) > len(self.particles):
new_particles.remove(self.least_likely(new_particles))
self.particles = new_particles
duration = time.time() - t0
