Survey: Human-Robot Collaboration through Vision-and-Language Navigation

Authors: Nivedan Yakolli, Avinash Gautam, Abhijit Das, Yuankai Qi, Virendra Singh Shekhawat Reviewed: ~200 articles

Overview

VLN (Vision-and-Language Navigation) is a multimodal task where agents interpret human language instructions to navigate 3D environments. This survey focuses on how VLN advances human-robot collaboration.

Key Gaps Identified

Current VLN/HRC systems struggle with:

  • Bidirectional communication — most systems only receive instructions, can’t ask clarifying questions
  • Ambiguity resolution — unclear instructions cause navigation failures with no recovery mechanism
  • Collaborative decision-making in multi-agent systems — multiple robots lack coordination frameworks

Recommendations

Systems should incorporate:

  • Proactive clarification mechanisms (robot asks when uncertain)
  • Real-time feedback capabilities
  • Contextual reasoning through advanced natural language understanding
  • Decentralized decision-making with dynamic role assignment

Application Domains

Healthcare, logistics, disaster response — environments where HRC failures have high stakes and bidirectional communication is essential.