FLEX: A Framework for Learning Robot-Agnostic Force-based Skills Involving Sustained Contact Object Manipulation

2025-03-17

AI Summary

Continuous contact-based manipulation tasks—such as pushing, sliding, and opening drawers/doors—exhibit poor generalization across unseen objects and diverse robotic platforms. Method: This paper proposes an object-centric, force-space-driven manipulation framework that decouples robot-body dynamics from object dynamics. Its core innovations include: (i) the first paradigm for region-oriented force application in force space; (ii) a robot-agnostic action representation coupled with a joint-configuration-aware dynamics model; and (iii) a pipeline integrating simulation pretraining, in-domain reinforcement learning, and kinematically decoupled control. Results: Experiments demonstrate over 10× improvement in policy training efficiency; zero-shot transfer to unseen objects and multiple real-world platforms (Kinova, Panda, UR5) with high success rates in both simulation and physical settings; and, for the first time, empirical validation of force-based skill generalization across heterogeneous robotic platforms on real hardware.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Learning to manipulate objects efficiently, particularly those involving sustained contact (e.g., pushing, sliding) and articulated parts (e.g., drawers, doors), presents significant challenges. Traditional methods, such as robot-centric reinforcement learning (RL), imitation learning, and hybrid techniques, require massive training and often struggle to generalize across different objects and robot platforms. We propose a novel framework for learning object-centric manipulation policies in force space, decoupling the robot from the object. By directly applying forces to selected regions of the object, our method simplifies the action space, reduces unnecessary exploration, and decreases simulation overhead. This approach, trained in simulation on a small set of representative objects, captures object dynamics -- such as joint configurations -- allowing policies to generalize effectively to new, unseen objects. Decoupling these policies from robot-specific dynamics enables direct transfer to different robotic platforms (e.g., Kinova, Panda, UR5) without retraining. Our evaluations demonstrate that the method significantly outperforms baselines, achieving over an order of magnitude improvement in training efficiency compared to other state-of-the-art methods. Additionally, operating in force space enhances policy transferability across diverse robot platforms and object types. We further showcase the applicability of our method in a real-world robotic setting. For supplementary materials and videos, please visit: https://tufts-ai-robotics-group.github.io/FLEX/

Problem

• Learning force-based object manipulation skills
• Generalizing across different objects and robots
• Reducing training time and simulation overhead

Innovation

• Object-centric manipulation in force space
• Decouples robot dynamics from object manipulation
• Enables direct transfer across robotic platforms