QUART-Online: Latency-Free Large Multimodal Language Model for Quadruped Robot Learning

TL;DR

This paper introduces QUART-Online, a latency-free multimodal language model for quadruped robots that achieves real-time inference and significantly improves task success rates by integrating vision, language, and compressed actions.

Contribution

The paper presents a novel latency-free MLLM model with Action Chunk Discretization, enabling real-time inference without performance loss during action instruction tuning.

Findings

Achieves real-time inference aligned with controller frequency.

Boosts task success rate by 65%.

Maintains language model performance despite latency reduction techniques.

Abstract

This paper addresses the inherent inference latency challenges associated with deploying multimodal large language models (MLLM) in quadruped vision-language-action (QUAR-VLA) tasks. Our investigation reveals that conventional parameter reduction techniques ultimately impair the performance of the language foundation model during the action instruction tuning phase, making them unsuitable for this purpose. We introduce a novel latency-free quadruped MLLM model, dubbed QUART-Online, designed to enhance inference efficiency without degrading the performance of the language foundation model. By incorporating Action Chunk Discretization (ACD), we compress the original action representation space, mapping continuous action values onto a smaller set of discrete representative vectors while preserving critical information. Subsequently, we fine-tune the MLLM to integrate vision, language, and compressed actions into a unified semantic space. Experimental results demonstrate that QUART-Online operates in tandem with the existing MLLM system, achieving real-time inference in sync with the underlying controller frequency, significantly boosting the success rate across various tasks by 65%. Our project page is https://quart-online.github.io.