Dynamic grid size required?

Hi everyone,

I just had a minor question. In some of the examples, the width and height of the environment are randomly sampled, while in others they are fixed. Is it a requirement that your algorithm should be able to handle dynamic grid sizes? Or can we assume a fixed grid size.

Thanks in advance!


Hello @tim_resink,

I am curious what examples you are referring to?

The detailed configurations of the environments used for evaluation, including their dimensions, are publicly known in Round 1:

As there are 14 different configurations it would makes sense that your algorithm handles arbitrary grid sizes (at least up to 150x150)!

Hi @MasterScrat,

Thank you for the quick reply, I hadn’t found the configuration page yet.

In the actual multi_agent_training.py file (which differs from the one shown on the RL–> multiple agents intro on the website), there is only a fixed size of 35x35. Here is the link:

1 Like

Thanks for the link. Indeed during training there are multiple strategies: either focus on a single configuration at a time, or make some sort of “curriculum” to make your agent more general!