Wrong steer in Datasets？

[Lzc6996]

Issue Description
I have noticed inconsistencies in the steer values while using the dataset. Here are some specific cases I observed:

1. In the file CrossingBicycleFlow_Town12_Route1063_Weather23/anno/00343.json.gz, the agent is turning, but the steer value is 0.
2. In the file CrossingBicycleFlow_Town12_Route862_Weather3/anno/00342.json.gz, the agent's steer value is -0.5, but in 00343 it changes to -0.1, and then jumps back a few frames later.
   This phenomenon seems to be prevalent throughout the dataset. Could this be an issue with the data itself, or am I misunderstanding something in my usage?

Thank you for your help!

[jayyoung0802]

Turning is a process, steer is an instantaneous value. Since the action space of the RL agent is discrete, the actions are not smooth. You can watch our demo to feel this process.