AstraNav-Memory: Contexts Compression for Long Memory

TL;DR

This paper introduces AstraNav-Memory, an image-centric memory compression framework for embodied navigation that enhances long-term memory capacity and improves navigation performance in diverse environments.

Contribution

It presents a novel visual context compression module integrated with a VL-based navigation policy, enabling scalable long-term memory for embodied agents.

Findings

Achieves state-of-the-art navigation performance on GOAT-Bench and HM3D-OVON.

Supports configurable compression rates, with 16× compression encoding about 30 tokens per image.

Moderate compression balances efficiency and accuracy effectively.

Abstract

Lifelong embodied navigation requires agents to accumulate, retain, and exploit spatial-semantic experience across tasks, enabling efficient exploration in novel environments and rapid goal reaching in familiar ones. While object-centric memory is interpretable, it depends on detection and reconstruction pipelines that limit robustness and scalability. We propose an image-centric memory framework that achieves long-term implicit memory via an efficient visual context compression module end-to-end coupled with a Qwen2.5-VL-based navigation policy. Built atop a ViT backbone with frozen DINOv3 features and lightweight PixelUnshuffle+Conv blocks, our visual tokenizer supports configurable compression rates; for example, under a representative 16$\times$ compression setting, each image is encoded with about 30 tokens, expanding the effective context capacity from tens to hundreds of images. Experimental results on GOAT-Bench and HM3D-OVON show that our method achieves state-of-the-art navigation performance, improving exploration in unfamiliar environments and shortening paths in familiar ones. Ablation studies further reveal that moderate compression provides the best balance between efficiency and accuracy. These findings position compressed image-centric memory as a practical and scalable interface for lifelong embodied agents, enabling them to reason over long visual histories and navigate with human-like efficiency.