Context-Aware Force Estimation for Deformable Tool Manipulation in Robotic Environmental Swabbing via Few-Shot Continual Adaptation

2026-07-08

AI Summary

This work addresses the challenge of accurate contact force estimation in robotic swabbing, where viscoelastic hysteresis in compliant tools compromises wrist-mounted force sensors and precludes sensor integration at the tool tip. To overcome this, the authors propose a proprioception-based, data-driven framework that leverages an LSTM to model the tool’s deformation history and employs a parameter-isolated Feature-wise Linear Modulation (FiLM) mechanism to generate low-dimensional context embeddings. This enables lightweight, few-shot continual cross-domain adaptation with a frozen backbone network, effectively decoupling shared deformation dynamics from task-specific context and mitigating catastrophic forgetting. Evaluated on a UR5e platform across nine tool–surface interaction scenarios, the method reduces zero-shot force estimation error by up to 63%, achieves inference latency under 1 ms, and preserves baseline performance.

Abstract

Robotic surface swabbing requires sustained interaction between a compliant tool and heterogeneous environments, where accurate estimation of tip-level contact force is critical for consistent sampling performance. However, deformable tool dynamics introduce nonlinear viscoelastic hysteresis that decouples wrist-mounted force measurements from true contact forces, while tool-integrated sensors are impractical for deployment due to sterility and disposability constraints. This paper presents a data-driven framework for contact force estimation in Deformable Tool Manipulation (DTM) that leverages proprioceptive sensing without requiring explicit physical models or permanent embedded sensing hardware at the tool tip. A recurrent architecture is first identified through a comparative evaluation of temporal models, where a compact LSTM achieves the lowest estimation error and sub-millisecond inference latency. To address generalization across unseen surfaces and tool compliance conditions, we introduce a parameter-isolated few-shot adaptation strategy that augments a frozen recurrent backbone with low-dimensional context embeddings using feature-wise linear modulation (FiLM). Experiments on a UR5e platform across nine tool-surface interaction regimes demonstrate that the proposed approach significantly improves robustness under domain shift, reducing zero-shot estimation error by up to 63\% while preserving baseline performance without catastrophic forgetting. These results show that separating shared deformation-history dynamics from domain-specific conditioning enables reliable force estimation for DTM in non-stationary environments.

Problem

• Deformable Tool Manipulation
• Contact Force Estimation
• Context-Aware Adaptation
• Robotic Swabbing
• Domain Generalization

Innovation

• few-shot continual adaptation
• context-aware force estimation
• deformable tool manipulation
• FiLM conditioning
• proprioceptive sensing