MCNav: Memory-Aware Dynamic Cognitive Map for Zero-shot Goal-oriented Navigation

2026-05-19

AI Summary

This work addresses the problem of target omission or misidentification in zero-shot object navigation, which often arises from neglecting information about previously explored regions. To mitigate this, the authors propose a memory-aware dynamic cognitive map that stores and retrieves object-related information from past exploration. Building upon this map, they introduce two memory-driven strategies—target re-verification and missed-target re-exploration—augmented with a blacklist mechanism and dual confirmation protocol to enhance navigation robustness. Evaluated on the HM3Dv1 and HM3Dv2 datasets across three benchmark tasks, the proposed method achieves state-of-the-art performance, demonstrating particularly significant improvements in instance-level object navigation compared to existing approaches.

Abstract

Navigating to instance-level targets in complex environments is a challenging problem. Many existing zero-shot methods achieve strong performance by modeling the entire environment and leveraging large language models for scene understanding. However, such strategies primarily focus on exploring new regions while lacking a deeper exploitation of information from previously explored areas. Consequently, when targets are missed or misidentified within previously visited regions, navigation failures occur frequently. To address these limitations, we propose MCNav, a memory-aware navigation framework with a dynamic cognitive map. This map stores efficiently queryable information about relevant objects in explored areas. Building on this memory structure, MCNav introduces two memory-aware exploration strategies: goal re-validation, which re-assesses previously seen objects to correct matching failures, and missed goal re-exploration, which estimates the likelihood that a target is present in an explored region from contextual cues. These strategies are further stabilized by a blacklist mechanism to prevent repeated errors and a double-check mechanism for high-confidence confirmation. We evaluate MCNav on the HM3Dv1 and HM3Dv2 datasets across three different tasks, where it achieves state-of-the-art performance, particularly on the instance-level goal navigation task.

Problem

• zero-shot navigation
• instance-level goal
• memory exploitation
• navigation failure
• explored regions

Innovation

• memory-aware navigation
• dynamic cognitive map
• zero-shot goal navigation
• goal re-validation
• missed goal re-exploration