Lightweight Learning from Actuation-Space Demonstrations via Flow Matching for Whole-Body Soft Robotic Grasping

2025-11-03

AI Summary

Soft-bodied robots face challenges in whole-body grasping under uncertainty and rich contact conditions, typically requiring dense sensing and complex closed-loop feedback control. To address this, we propose a lightweight actuation-space learning framework. Our approach innovatively introduces Rectified Flow—a flow-matching generative model—into actuation-space modeling, enabling direct learning of distributed, probabilistic control representations from only 30 deterministic demonstrations, without explicit dynamics models or real-time feedback. By leveraging the inherent passive adaptability of soft structures as functional control intelligence, the method significantly reduces computational load on the central controller. Experiments demonstrate a 97.5% grasping success rate and strong generalization to ±33% object size variations and 20%–200% execution time scaling. The framework achieves end-to-end control with minimal samples, low sensory requirements, and high robustness.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Robotic grasping under uncertainty remains a fundamental challenge due to its uncertain and contact-rich nature. Traditional rigid robotic hands, with limited degrees of freedom and compliance, rely on complex model-based and heavy feedback controllers to manage such interactions. Soft robots, by contrast, exhibit embodied mechanical intelligence: their underactuated structures and passive flexibility of their whole body, naturally accommodate uncertain contacts and enable adaptive behaviors. To harness this capability, we propose a lightweight actuation-space learning framework that infers distributional control representations for whole-body soft robotic grasping, directly from deterministic demonstrations using a flow matching model (Rectified Flow),without requiring dense sensing or heavy control loops. Using only 30 demonstrations (less than 8% of the reachable workspace), the learned policy achieves a 97.5% grasp success rate across the whole workspace, generalizes to grasped-object size variations of +-33%, and maintains stable performance when the robot's dynamic response is directly adjusted by scaling the execution time from 20% to 200%. These results demonstrate that actuation-space learning, by leveraging its passive redundant DOFs and flexibility, converts the body's mechanics into functional control intelligence and substantially reduces the burden on central controllers for this uncertain-rich task.

Problem

• Learning actuation-space control for soft robotic grasping
• Addressing uncertainty in contact-rich grasping tasks
• Reducing controller burden via passive mechanical intelligence

Innovation

• Flow matching model learns from actuation-space demonstrations
• Lightweight framework reduces need for dense sensing
• Converts robot body mechanics into control intelligence