OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints

TL;DR

OmniManip introduces an object-centric approach that leverages spatial constraints and interaction primitives to enable general robotic manipulation without fine-tuning Vision-Language Models, achieving strong zero-shot generalization.

Contribution

The paper presents a novel object-centric representation and a dual closed-loop system that bridges high-level reasoning and low-level control for general manipulation tasks.

Findings

Strong zero-shot generalization across diverse tasks

Robust real-time control without VLM fine-tuning

Effective automation of large-scale simulation data generation

Abstract

The development of general robotic systems capable of manipulating in unstructured environments is a significant challenge. While Vision-Language Models(VLM) excel in high-level commonsense reasoning, they lack the fine-grained 3D spatial understanding required for precise manipulation tasks. Fine-tuning VLM on robotic datasets to create Vision-Language-Action Models(VLA) is a potential solution, but it is hindered by high data collection costs and generalization issues. To address these challenges, we propose a novel object-centric representation that bridges the gap between VLM's high-level reasoning and the low-level precision required for manipulation. Our key insight is that an object's canonical space, defined by its functional affordances, provides a structured and semantically meaningful way to describe interaction primitives, such as points and directions. These primitives act as a bridge, translating VLM's commonsense reasoning into actionable 3D spatial constraints. In this context, we introduce a dual closed-loop, open-vocabulary robotic manipulation system: one loop for high-level planning through primitive resampling, interaction rendering and VLM checking, and another for low-level execution via 6D pose tracking. This design ensures robust, real-time control without requiring VLM fine-tuning. Extensive experiments demonstrate strong zero-shot generalization across diverse robotic manipulation tasks, highlighting the potential of this approach for automating large-scale simulation data generation.