Gondola: Grounded Vision Language Planning for Generalizable Robotic Manipulation
This work addresses the problem of precise object grounding in robotic manipulation for researchers and practitioners, offering a novel method that improves generalization but is incremental in combining multi-view inputs with LLMs.
The paper tackles the challenge of generalizing robotic manipulation across unseen objects, environments, and tasks by introducing Gondola, a grounded vision-language planning model that uses multi-view images and history plans to generate action plans with segmentation masks, achieving state-of-the-art performance on the GemBench dataset across all generalization levels.
Robotic manipulation faces a significant challenge in generalizing across unseen objects, environments and tasks specified by diverse language instructions. To improve generalization capabilities, recent research has incorporated large language models (LLMs) for planning and action execution. While promising, these methods often fall short in generating grounded plans in visual environments. Although efforts have been made to perform visual instructional tuning on LLMs for robotic manipulation, existing methods are typically constrained by single-view image input and struggle with precise object grounding. In this work, we introduce Gondola, a novel grounded vision-language planning model based on LLMs for generalizable robotic manipulation. Gondola takes multi-view images and history plans to produce the next action plan with interleaved texts and segmentation masks of target objects and locations. To support the training of Gondola, we construct three types of datasets using the RLBench simulator, namely robot grounded planning, multi-view referring expression and pseudo long-horizon task datasets. Gondola outperforms the state-of-the-art LLM-based method across all four generalization levels of the GemBench dataset, including novel placements, rigid objects, articulated objects and long-horizon tasks.