Spatial RoboGrasp: Generalized Robotic Grasping Control Policy

TL;DR

This paper introduces a unified spatial perception and diffusion-based policy framework that significantly improves robotic grasping success rates across diverse environments by integrating multimodal perception and task prompts.

Contribution

It presents a novel framework combining domain augmentation, depth estimation, and diffusion models for robust, generalizable robotic grasping under varied conditions.

Findings

Up to 40% improvement in grasp success rate

45% higher task success rate under environmental variation

Effective integration of multimodal perception with diffusion policies

Abstract

Achieving generalizable and precise robotic manipulation across diverse environments remains a critical challenge, largely due to limitations in spatial perception. While prior imitation-learning approaches have made progress, their reliance on raw RGB inputs and handcrafted features often leads to overfitting and poor 3D reasoning under varied lighting, occlusion, and object conditions. In this paper, we propose a unified framework that couples robust multimodal perception with reliable grasp prediction. Our architecture fuses domain-randomized augmentation, monocular depth estimation, and a depth-aware 6-DoF Grasp Prompt into a single spatial representation for downstream action planning. Conditioned on this encoding and a high-level task prompt, our diffusion-based policy yields precise action sequences, achieving up to 40% improvement in grasp success and 45% higher task success rates under environmental variation. These results demonstrate that spatially grounded perception, paired with diffusion-based imitation learning, offers a scalable and robust solution for general-purpose robotic grasping.