LERR-Nav: Learning to Coordinate Exploration, Recovery, and Reminiscing for Object Navigation

Key points

• Object Navigation in unseen environments remains a fundamental challenge in embodied AI, as it requires agents to generalize visual and spatial reasoning to novel layouts.1
• While recent advances in large language model (LLM) and vision-language model (VLM) have enhanced semantic mapping, spatial memory, and waypoint selecting for navigation, successful navigation requires more than selecting the next promising waypoint.1
• To support history-aware coordination, we maintain the execution memory that preserves textual summaries of all strategy executions alongside a sliding window of map sequences.1
• The second stage performs SFT on variable-length navigation segments to learn which navigation strategy to invoke next based on current maps and the hybrid execution memory.1

Sources (1)

• [1]LERR-Nav: Learning to Coordinate Exploration, Recovery, and Reminiscing for Object Navigation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 05:38 AM

  Motivated by the analysis of failed navigation episodes, we propose LERR-Nav, an object navigation framework that coordinates exploration, recovery, and reminiscing through a learned high-level policy.

  Object Navigation in unseen environments remains a fundamental challenge in embodied AI, as it requires agents to generalize visual and spatial reasoning to novel layouts.

Extractive summary: sentences quoted from the sources.