K-VARK: Kernelized Variance-Aware Residual Kalman Filter for Sensorless Force Estimation in Collaborative Robots

2025-12-15

AI Summary

To address insufficient accuracy and robustness in sensorless contact force estimation for collaborative robots—caused by modeling errors, residual dynamics, and nonlinear friction—this paper proposes a novel framework integrating Kernelized Movement Primitives (KMP) with adaptive Kalman filtering. Specifically, we employ a heteroscedastic Gaussian process to jointly model the mean and input-dependent noise of residual torque. A variance-aware virtual measurement update mechanism is designed, and variational Bayesian inference is introduced for online optimization of the process noise covariance. Experimental validation on a 6-DOF collaborative manipulator demonstrates that the proposed method reduces root-mean-square error (RMSE) by over 20% compared to the state-of-the-art sensorless approaches. This improvement significantly enhances the reliability and adaptability of force control in practical tasks such as robotic polishing and precision assembly.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Kernel Methods
• Natural Language Processing: Safety and Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Reliable estimation of contact forces is crucial for ensuring safe and precise interaction of robots with unstructured environments. However, accurate sensorless force estimation remains challenging due to inherent modeling errors and complex residual dynamics and friction. To address this challenge, in this paper, we propose K-VARK (Kernelized Variance-Aware Residual Kalman filter), a novel approach that integrates a kernelized, probabilistic model of joint residual torques into an adaptive Kalman filter framework. Through Kernelized Movement Primitives trained on optimized excitation trajectories, K-VARK captures both the predictive mean and input-dependent heteroscedastic variance of residual torques, reflecting data variability and distance-to-training effects. These statistics inform a variance-aware virtual measurement update by augmenting the measurement noise covariance, while the process noise covariance adapts online via variational Bayesian optimization to handle dynamic disturbances. Experimental validation on a 6-DoF collaborative manipulator demonstrates that K-VARK achieves over 20% reduction in RMSE compared to state-of-the-art sensorless force estimation methods, yielding robust and accurate external force/torque estimation suitable for advanced tasks such as polishing and assembly.

Problem

• Estimates contact forces without sensors in robots
• Addresses modeling errors and complex residual dynamics
• Improves accuracy for safe robot-environment interaction

Innovation

• Kernelized probabilistic model of joint residual torques
• Variance-aware virtual measurement update via noise covariance augmentation
• Online adaptive process noise via variational Bayesian optimization