Translating Natural Language Instructions for Behavioral Robot Navigation with a Multi-Head Attention Mechanism

We propose a multi-head attention mechanism as a blending layer in a neural\nnetwork model that translates natural language to a high level behavioral\nlanguage for indoor robot navigation. We follow the framework established by\n(Zang et al., 2018a) that proposes the use of a navigation graph as a knowledge\nbase for the task. Our results show significant performance gains when\ntranslating instructions on previously unseen environments, therefore,\nimproving the generalization capabilities of the model.\n