dqn_drone.py returns error: Observation Space must be an image or dictionary observation space

[shashwatkansal]

Bug report

• AirSim Version/#commit: 052228a
• UE/Unity version: 5.1.1
• autopilot version:
• OS Version: Windows11 Version 22H2, OS Build 22621.1702

What's the issue you encountered?

I have built AirSim, I run the environments Blocks (or any other environment such as MountainLandscape), and upon running, then I navigate in terminal and run in a python pipenv (Python 3.9.12) PythonClient/reinforcement_learning/dqn_drone.py. The DQN starts off and very shortly thereafter returns an error

<stable_baselines3.common.vec_env.dummy_vec_env.DummyVecEnv object at 0x000001EAC077CFD0>
Traceback (most recent call last):
  File "C:\Users\shash\Documents\Thesis\Colosseum\PythonClient\reinforcement_learning\dqn_drone.py", line 28, in <module>
    env = VecTransposeImage(env)
  File "C:\Users\shash\.virtualenvs\PythonClient-mo3ke7LZ\lib\site-packages\stable_baselines3\common\vec_env\vec_transpose.py", line 22, in __init__
    assert is_image_space(venv.observation_space) or isinstance(
AssertionError: The observation space must be an image or dictionary observation space

This is purely based on default files without edits. Unsure how to go about fixing this?

[BetterLL]

oh I have the same question, how do you solve it?