MemER: Scaling Up Memory for Robot Control via Experience Retrieval

TL;DR

MemER introduces a hierarchical memory-based policy framework for robots, enabling efficient long-term reasoning and outperforming prior methods in real-world manipulation tasks by selectively retrieving relevant past experiences.

Contribution

The paper presents a hierarchical policy system that incorporates experience retrieval for robotic control, compatible with existing vision-language models, and demonstrates improved performance on long-horizon tasks.

Findings

Outperforms prior methods on three robotic tasks

Enables efficient long-term memory reasoning

Compatible with existing vision-language-action models

Abstract

Humans routinely rely on memory to perform tasks, yet most robot policies lack this capability; our goal is to endow robot policies with the same ability. Naively conditioning on long observation histories is computationally expensive and brittle under covariate shift, while indiscriminate subsampling of history leads to irrelevant or redundant information. We propose a hierarchical policy framework, where the high-level policy is trained to select and track previous relevant keyframes from its experience. The high-level policy uses selected keyframes and the most recent frames when generating text instructions for a low-level policy to execute. This design is compatible with existing vision-language-action (VLA) models and enables the system to efficiently reason over long-horizon dependencies. In our experiments, we finetune Qwen2.5-VL-7B-Instruct and $\pi_{0.5}$ as the high-level and low-level policies respectively, using demonstrations supplemented with minimal language annotations. Our approach, MemER, outperforms prior methods on three real-world long-horizon robotic manipulation tasks that require minutes of memory. Videos and code can be found at https://jen-pan.github.io/memer/.