QUART-Online: Latency-Free Large Multimodal Language Model for Quadruped Robot Learning
It addresses latency issues for quadruped robot learning, enabling more efficient and effective vision-language-action integration, though it appears incremental as it builds on existing MLLM systems.
This paper tackles the inference latency problem in deploying multimodal large language models for quadruped robot tasks by introducing QUART-Online, which uses Action Chunk Discretization to compress actions and fine-tune the model, achieving a 65% boost in success rates while enabling real-time inference.
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.