Imitation learning

[menichel]

Hi all

I am very interested in using your library for coordinating autonomous furniture in a room. What attracts me is that it seems that your agents always find a solution to reach the goal by avoiding local minima and collisions. For practical reasons I wanted to ask if it is necessary to do imitation learning and retrain the NN for each scenario or do you already provide a trained NN that is able to tackle most of the scenarios?

Thanks for your help and have a nice day

menichel