# Action Space

The action space of every `RoarRLEnv` would be identical to the return value of `RoarPyActor.get_action_spec()`

{% hint style="info" %}
See [here](https://github.com/augcog/ROAR_PY/blob/e9ba4e514016b0d3a2bf35972b38633b2b94b741/roar_py_carla/roar_py_carla/actors/carla_vehicle.py#L20) for the action space for Carla vehicles
{% endhint %}

{% hint style="info" %}
For ROAR group's internal RL code, we had to filter out some action dimensions to make training quicker. See the following link for how we filter out actions.\
<https://github.com/augcog/ROAR_PY_RL/blob/e23f221d983bd7fb8c39d86af27828f6f184447b/training/env_util.py#L9>
{% endhint %}