LidarObservation Warning - divide by zero

[Rock3Yu]

I face the same problem again. Observation: Lidar, Env: parking-v0

version:

Name: highway-env
Version: 1.10.1

I change the env config of parking-v0 with only observation and vehicles_count=0/3/10:

config["observation"] = {
    "type": "LidarObservation",
    "cells": 128,
    "maximum_range": 64,
    "normalise": True,
}

It will only happen when obs=lidar. I tested obs=default/GrayscaleObservation, both are good. (use stable_baselines3)

The UserWarning/RuntimeWarning is like:

/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/gymnasium/utils/passive_env_checker.py:134: UserWarning: WARN: The obs returned by the `reset()` method was expecting numpy array dtype to be float32, actual type: float64
  logger.warn(
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/gymnasium/utils/passive_env_checker.py:158: UserWarning: WARN: The obs returned by the `reset()` method is not within the observation space.
  logger.warn(f"{pre} is not within the observation space.")
Using cpu device
Wrapping the env with a `Monitor` wrapper
Wrapping the env in a DummyVecEnv.
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
Logging to ./tensorboard_logs/lidar_PPO_empty/PPO_1
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/gymnasium/utils/passive_env_checker.py:134: UserWarning: WARN: The obs returned by the `step()` method was expecting numpy array dtype to be float32, actual type: float64
  logger.warn(
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/gymnasium/utils/passive_env_checker.py:158: UserWarning: WARN: The obs returned by the `step()` method is not within the observation space.
  logger.warn(f"{pre} is not within the observation space.")
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
/opt/anaconda3/envs/yu/lib/python3.9/site-packages/highway_env/utils.py:403: RuntimeWarning: divide by zero encountered in scalar divide
  interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]

Originally posted by @Rock3Yu in #650 (comment)

[JoepdeBoer]

The algorithm is not correctly implemented , if the array is parallel to one side of the object (rectangle) the dot product wil be zero. A quick fix is this although not thoroughly checked. However i believe there must be a cleaner solution out there.
if rqu==0:
inf1 = copysign(np.inf, (a - r) @ u)
inf2 = copysign(np.inf, (b - r) @ u)
interval_1 = [inf1,inf2]
interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
interval_2 = interval_2 if rqv >= 0 else list(reversed(interval_2))
elif rqv == 0:
inf1 = copysign(np.inf, (a - r) @ v)
inf2 = copysign(np.inf, (d - r) @ v)
interval_2 = [inf1, inf2] # TODO check if sign needs to be correct
interval_1 = [(a - r) @ u / rqu, (b - r) @ u / rqu]
interval_1 = interval_1 if rqu >= 0 else list(reversed(interval_1))

else:
    interval_1 = [(a - r) @ u / rqu, (b - r) @ u / rqu]
    interval_1 = interval_1 if rqu >= 0 else list(reversed(interval_1))
    interval_2 = [(a - r) @ v / rqv, (d - r) @ v / rqv]
    interval_2 = interval_2 if rqv >= 0 else list(reversed(interval_2))