Context vector-based visual mapless navigation in indoor using hierarchical semantic information and meta-learning

Abstract Visual mapless navigation (VMN), modeling a direct mapping between sensory inputs and agent actions, aims to navigate from stochastic origin location prescribed goal in an unseen scene. A fundamental yet challenging issue visual is generalizing new Furthermore, it of pivotal concern design method make effective policy learning. To address these issues, we introduce novel model, which integrates hierarchical semantic information represented by context vector with meta-learning improve the generalization performance gap known unknown environments. Extensive experimental results on AI2-THOR benchmark dataset demonstrate that our model significantly outperforms state-of-the-art $$15.79\%$$ 15.79 % for SPL $$23.83\%$$ 23.83 success rate. In addition, exploration rate experiment shows can effectively invalid behavior accelerate convergence speed model. Our implementation code data be viewed https://github.com/zhiyu-tech/WHU-CVVMN .