How an agent can change its brain?

[pooyanjamshidi]

In order to have a flexible cooperative architecture (e.g., Mergeable nervous systems for robots, Nature 2017), the ability to chain the brain, at any point, seems essential. I was wondering whether it is possible for the agents to change their brain at runtime? Also, is it possible to form a structure that can be changed at runtime for cooperative tasks?

[awjuliani]

Hi @pooyanjamshidi,

Thanks for your interest in ML-Agents. We've built in the ability to change brains during training. An agent can call its GiveBrain() function to replace the current brain with a different one.

I am curious what you mean by structure that can be changed? Does this apply to the structure of the agent itself, or the environment?

[pooyanjamshidi]

Hi @awjuliani,

Thanks for your prompt reply.

In the cooperative scenario, I assume several agents can interact to finish a task. In this scenario, one or several agents may fail for whatever reason. How might new agents be recruited in order to join the other agents for finishing the task? More specifically, is there any mechanisms through which we could define a structure (e.g., delegating sub-tasks) and change the structure over time, many examples of which can be seen in nature, the way birds travels, or fishes go for hunting, etc.

[awjuliani]

Hi @pooyanjamshidi, You can change the structure of the problem dynamically, so long as the state and action spaces remain fixed.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.