the form of input data

[XuChang-c]

Hi, can you tell me the form of the observation, such as vehicle. Is it speed and location?

[Luca96]

Hi,
you can find the description of the observation space here. In particular:

• image: is a concatenation of tree RGB camera images, along width axis.
• road: see the link.
• vehicle: comprises vehicle's speed, throttle, brake, and (cosine) similarity with the road waypoint.
• navigation: includes next road locations that the agent has to reach.

The actual position of the vehicle is not included in the observation space, but you can retrieve it by calling get_location() from a vehicle instance: you find it in CARLAEnv for example.