Varying Agent size/scale?

Hi, for the wall jump demo is it possible to train using a randomized scale? So that the agent is a random height/width for the jump attempts? Would something like this be possible? I'm brand new to ML stuff.

Thanks!

 

I think you could, but you will need to somehow feed the scale of the agent as observation. Also, I do not know how scale would impact the performance of the agent. It would probably change the height of the raycasts, which might make it harder to train.