Implicit Neural-Representation Learning for Elastic Deformable-Object Manipulations

2025-05-01

AI Summary

Manipulating elastic deformable objects (e.g., rubber bands) in real-world settings remains challenging due to their infinite-dimensional state space and sparse, partial observations (e.g., RGB-D images or point clouds), which impede accurate state estimation and robust policy learning. To address this, we propose INR-RL: the first framework that jointly optimizes implicit neural representations—specifically signed distance functions (SDFs)—with deep reinforcement learning (PPO/SAC) in an end-to-end manner, enabling simultaneous mapping from partial observations to continuous geometric representations and policy training. We further enhance representation consistency via self-supervised reconstruction and exploration-driven fine-tuning. Evaluated on a Franka Panda robot via sim-to-real transfer, INR-RL significantly improves deformation state estimation accuracy and manipulation success rates, while generalizing effectively to unseen configurations and novel elastic materials.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments

Abstract

We aim to solve the problem of manipulating deformable objects, particularly elastic bands, in real-world scenarios. However, deformable object manipulation (DOM) requires a policy that works on a large state space due to the unlimited degree of freedom (DoF) of deformable objects. Further, their dense but partial observations (e.g., images or point clouds) may increase the sampling complexity and uncertainty in policy learning. To figure it out, we propose a novel implicit neural-representation (INR) learning for elastic DOMs, called INR-DOM. Our method learns consistent state representations associated with partially observable elastic objects reconstructing a complete and implicit surface represented as a signed distance function. Furthermore, we perform exploratory representation fine-tuning through reinforcement learning (RL) that enables RL algorithms to effectively learn exploitable representations while efficiently obtaining a DOM policy. We perform quantitative and qualitative analyses building three simulated environments and real-world manipulation studies with a Franka Emika Panda arm. Videos are available at http://inr-dom.github.io.

Problem

• Manipulating elastic deformable objects with partial observations
• Addressing high-dimensional state space in deformable object manipulation
• Learning consistent implicit representations for effective policy training

Innovation

• Implicit neural-representation learning for deformable objects
• Reinforcement learning fine-tunes state representations
• Signed distance function reconstructs complete object surfaces