TDANet: Target-Directed Attention Network For Object-Goal Visual Navigation With Zero-Shot Ability

arXiv:2404.08353v1 Announce Type: new
Abstract: The generalization of the end-to-end deep reinforcement learning (DRL) for object-goal visual navigation is a long-standing challenge since object classes and placements vary in new test environments. Learning domain-independent visual representation is critical for enabling the trained DRL agent with the ability to generalize to unseen scenes and objects. In this letter, a target-directed attention network (TDANet) is proposed to learn the end-to-end object-goal visual navigation policy with zero-shot ability. TDANet features a novel target …

abstract agent arxiv attention challenge cs.cv cs.ro domain enabling environments independent navigation network object reinforcement reinforcement learning representation test the end type visual visual navigation zero-shot