Auxiliary Tasks and Exploration Enable ObjectNav
Published in
arXiv.org(2020)External Links:
TL
TL;DR
It is proposed that agents will act to simplify their visual inputs so as to smooth their RNN dynamics, and that auxiliary tasks reduce overfitting by minimizing effective RNN dimensionality; i.e. a performant ObjectNav agent that must maintain coherent plans over long horizons does so by learning smooth, low-dimensional recurrent dynamics.