Documentation for AnimalAIEnvironment Class

[ehddnr747]

I am trying to spawn an environment and test/implement my own networks.

Following this, there are lots of parameters for AnimalAIEnvironment. I have no idea what each parameter does, and I cannot find any detailed documentation. It would be great help, if we have it.

[mdcrosby]

Hi. Most of the parameters are inherited from ml-agents and are documented by ml-agents. See here and related links. The remaining few are for the agent observations specific to AnimalAI and are documented under observations here.

[ehddnr747]

Thanks for your fast response. I remember that there was an option for creating multiple arenas in a simulation, n_arenas, but I cannot find it anymore in version 3.0. Would you kindly help me to spawn multiple arenas? Just letting me know any resources that I can refer to would be fine.

[mdcrosby]

This feature was removed. In testing it turned out to be more efficient to spawn multiple separate environment processes compared to having multiple arenas within a single environment. There are some additional startup costs this way, but those are minimal in the long-term and you gain more flexibility over each (sub)agent.