So looking at the videos it seems like at the second case the agent have almost doubled time to run..
The questions are:
– why videos have different length ?
– is it possible that one agent have more time than another. Am I right that all the agents act at max for N frames, where N = 30 frames * 16.6 seconds
I’m not sure what you mean by 16.6 seconds? Every agent has a set number of frames (timesteps) to run. The amount of “time” you get is dependent on if you stay alive for long enough, which is why some videos are longer than others.
No one gets more time – its mostly dependent on how well your agent does (if you drive off the edge for example, we end the episode short), but we also have a hard cap on the number of time steps.
for n in range(MAX_ENVIRONMENT_TIMESTEPS):
—- if done: # given from env