Mem2Ego: Empowering Vision-Language Models with Global-to-Ego Memory for Long-Horizon Embodied Navigation

TL;DR

Mem2Ego introduces a novel framework that combines global memory retrieval with egocentric visual inputs, significantly improving spatial reasoning and decision-making in long-horizon embodied navigation tasks.

Contribution

The paper presents a new vision-language navigation approach that adaptively integrates global memory with local observations, addressing limitations of previous methods in complex environments.

Findings

Outperforms previous state-of-the-art in object navigation tasks

Enhances spatial reasoning through global-to-ego memory integration

Demonstrates scalability and effectiveness in complex environments

Abstract

Recent advancements in Large Language Models (LLMs) and Vision-Language Models (VLMs) have made them powerful tools in embodied navigation, enabling agents to leverage commonsense and spatial reasoning for efficient exploration in unfamiliar environments. Existing LLM-based approaches convert global memory, such as semantic or topological maps, into language descriptions to guide navigation. While this improves efficiency and reduces redundant exploration, the loss of geometric information in language-based representations hinders spatial reasoning, especially in intricate environments. To address this, VLM-based approaches directly process ego-centric visual inputs to select optimal directions for exploration. However, relying solely on a first-person perspective makes navigation a partially observed decision-making problem, leading to suboptimal decisions in complex environments. In this paper, we present a novel vision-language model (VLM)-based navigation framework that addresses these challenges by adaptively retrieving task-relevant cues from a global memory module and integrating them with the agent's egocentric observations. By dynamically aligning global contextual information with local perception, our approach enhances spatial reasoning and decision-making in long-horizon tasks. Experimental results demonstrate that the proposed method surpasses previous state-of-the-art approaches in object navigation tasks, providing a more effective and scalable solution for embodied navigation.