def calculate_score_for_leaderboard():
"""
Evaluate the performance of the network. This is the function to be used for
the final ranking on the course-wide leader-board, only with a different set
of seeds. Better not change it.
"""
# action variables
a = np.array([0.0, 0.0, 0.0])
# init environement
env = CarRacing()
env.render()
env.reset()
seeds = [
22597174, 68545857, 75568192, 91140053, 86018367, 49636746, 66759182,
91294619, 84274995, 31531469
]
total_reward = 0
for episode in range(10):
env.seed(seeds[episode])
observation = env.reset()
# init modules of the pipeline
LD_module = LaneDetection(gradient_threshold=25, spline_smoothness=20)
LatC_module = LateralController(gain_constant=1.8,
damping_constant=0.05)
LongC_module = LongitudinalController(KD=0.001)
reward_per_episode = 0
for t in range(600):
# perform step
s, r, done, speed, info = env.step(a)
# lane detection
lane1, lane2 = LD_module.lane_detection(s)
# waypoint and target_speed prediction
waypoints = waypoint_prediction(lane1, lane2)
target_speed = target_speed_prediction(waypoints,
max_speed=60,
exp_constant=6)
# control
a[0] = LatC_module.stanley(waypoints, speed)
a[1], a[2] = LongC_module.control(speed, target_speed)
# reward
reward_per_episode += r
env.render()
print('episode %d \t reward %f' % (episode, reward_per_episode))
total_reward += np.clip(reward_per_episode, 0, np.infty)
print('---------------------------')
print(' total score: %f' % (total_reward / 10))
print('---------------------------')
# define variables
total_reward = 0.0
steps = 0
restart = False
# init modules of the pipeline
LD_module = LaneDetection()
LatC_module = LateralController()
LongC_module = LongitudinalController()
# init extra plot
fig = plt.figure()
plt.ion()
plt.show()
env.seed(91294619)
while True:
# perform step
s, r, done, speed, info = env.step(a)
# lane detection
lane1, lane2 = LD_module.lane_detection(s)
# waypoint and target_speed prediction
waypoints = waypoint_prediction(lane1, lane2)
target_speed = target_speed_prediction(waypoints,
max_speed=60,
exp_constant=4.5)
# control
a[0] = LatC_module.stanley(waypoints, speed)
if k == key.DOWN: a[2] = +0.8 # set 1.0 for wheels to block to zero rotation
def key_release(k, mod):
if k == key.LEFT and a[0] == -1.0: a[0] = 0
if k == key.RIGHT and a[0] == +1.0: a[0] = 0
if k == key.UP: a[1] = 0
if k == key.DOWN: a[2] = 0
env = CarRacing()
env.render()
env.viewer.window.on_key_press = key_press
env.viewer.window.on_key_release = key_release
while True:
env.seed(seed=5)
env.reset()
total_reward = 0.0
steps = 0
restart = False
env.render()
while True:
s, r, done, info = env.step(a)
total_reward += r
if steps % 200 == 0 or done:
# print("\naction " + str(["{:+0.2f}".format(x) for x in a]))
# print("step {} total_reward {:+0.2f}".format(steps, total_reward))
pass
steps += 1
env.render()
obs = state_to_1_batch_tensor(s)
env.reset()
# define variables
total_reward = 0.0
steps = 0
restart = False
# init modules of the pipeline
LD_module = LaneDetection()
LatC_module = LateralController()
# init extra plot
fig = plt.figure()
plt.ion()
plt.show()
env.seed(68545857)
while True:
# perform step
s, r, done, speed, info = env.step(a)
# lane detection
lane1, lane2 = LD_module.lane_detection(s)
# waypoint and target_speed prediction
waypoints = waypoint_prediction(lane1, lane2)
target_speed = target_speed_prediction(waypoints)
# control with constant gas and no braking
a[0] = LatC_module.stanley(waypoints, speed)