A Tactile Sensor-Based EKF Estimator to Predict Object-Contact States During Whole-Body MultiContact Manipulation

RAJESH SUBBURAMAN, Francesco Rachiglia, Olivier Stasse

Keywords: Tactile sensors, multi-sensor fusion, manipulation, multicontact, humanoid

TL;DR: An extended Kalman filter framework that uses whole-body tactile and sensor data to estimate object and contact states in real time for stable humanoid manipulation of large objects.

Abstract: We present an extended Kalman filter–based framework for online estimation of object and contact states during whole-body multi-contact manipulation of large objects with humanoids. By combining tactile skin patches, joint encoders, forward kinematics, and an IMU, the estimator tracks object and contact pose, twist, and forces in real time. Unlike conventional end-effector-centric pose estimation approaches, our approach leverages whole-body contact data available from skin patches to estimate the states of large and heavy objects during dynamic manipulation. Experiments on the TALOS humanoid performing dynamic grasp and lift tasks demonstrate promising results on tracking of object-contact states, with potential to enhance manipulation stability, and support complex whole-body manipulations including contact shuffling and can further broaden humanoid applications.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 22