Vector & Visual Observations?

Hi,

I've tried to find a clear answer to this but I haven't so far. Is it possible to provide the agent with vector observations and visual observations at the same time? It seems there have been some PRs about this on Github but I can't see it mentioned anywhere in the documentation so I'm wondering if it works?

Thanks

 

Yes, that's possible. All observations are collected in the context of sensors. So for vector observations, there is a dedicated vector sensor. It is being generated automatically if you set the vector observation size to any value > 0 in the behavior inspector. For using other sensors, like visual observations or raycasts, you need to add the corresponding components to your agent.

 

Thanks for the reply! Ok, I see, so it's done automatically there's nothing to actually configure on my side? Just add the observations I want in my code, set the observation size iii the behaviour inspector to however values I'm logging and then also have my camera sensor set up?

 

Yes, pretty much.
(You'll notice that if you set the vector observation size to 0, and then try to add observations to the sensor, it'll throw a null reference exception.)

 

Hey! So I've tried doing this but for some reason I'm getting the following error:

Fewer observations (0) made than vector observation size (13). The observations will be padded.

I can't really understand why this is happening as I've written it like they do in their example code (I've attached a screenshot). I've made sure that the useVecObs is ticked (ie true) in my component box... Any idea what could be going on here? Thanks