Ego doesn't stop when traffic lights are red or yellow

[MingfeiCheng]

I test Pylot by using my predefined scenario of CARLA and my e2e.conf is shown below:

--random_seed=1337
###### Object detection config ######
--simulator_obstacle_detection
###### Traffic light detector config ######
--simulator_traffic_light_detection
###### Depth #####
--perfect_depth_estimation
###### Segmentation config ######
--perfect_segmentation
###### Prediction config ######
--perfect_obstacle_tracking
--tracking_num_steps=10
--prediction
--prediction_num_past_steps=10
--prediction_num_future_steps=30
###### Planning config #####
--planning_type=waypoint
--prediction_type=linear
--control=pid
###### Control config #####
--steer_gain=1.0
--stop_for_people=True
--stop_for_traffic_lights=True
--stop_for_vehicles=True
--stop_at_uncontrolled_junctions=True
######### Logging config #########
--log_file_name=pylot.log
--csv_log_file_name=pylot.csv
--v=1
--visualize_rgb_camera

I find that the ego does not stop when traffic lights are red:

I think the method below might ignore the situation where traffic lights are red or yellow.

pylot/pylot/planning/world.py

Line 329 in 54eda46

def stop_traffic_light(self, tl, wp_vector, wp_angle) -> float:

My simple solution is modifying the code in L440

pylot/pylot/planning/world.py

Line 440 in 54eda46

return True, speed_factor_tl

The modification is shown below, which is useful for my testing scenarios.

if (tl.state == TrafficLightColor.RED or tl.state == TrafficLightColor.YELLOW):
        return True, 0
else:
        return True, speed_factor_tl

Is there a better solution for this problem?