Contact Observer for Humanoid Robot Pepper based on Tracking Joint Position Discrepancies

Abstract

In order to enable efficient control of a humanhumanoid in physical contact settings, a real-time solution for a contact observer is required. We propose a novel approach for proprioceptive sensor based contact sensing suitable for affordable personal robots with no force/torque or electric current sensing. We combine robot model knowledge and the output of acceleration resolved quadratic programming wholebody controller to make a prediction of expected position tracking error for computing our proposed contact observer signal. We demonstrate the efficiency of our approach in the experiments of contact detection and estimation of collision direction and intensity on a real humanoid robot Pepper platform controlled by a task-space multi-objective quadratic programming controller.

Multi-Contact Planning on Humans for Physical Assistance by Humanoid

“…Thus, the robot must be controlled to reach the person and established the contacts in closed-loop. In the continuation of our work, the robot's configuration, that is computed by our PG, will serve as a target for the closed-loop QP [32] controller, that will achieve the desired contacts and postures at best using online perception and measured contacts [33], [34]. In our controller, contacts will be made using guarded motion to absorb surface uncertainties.…”

Compliant Robot Motion Regulated via Proprioceptive Sensor Based Contact Observer

“…First we review the idea behind our methodology in [4]; then we present the details of the updated framework we propose in order to enhance both robustness and performance of our initial method for the case of multi-joint motions.…”

“…Recent developments in the area of proprioceptive sensor based contact monitoring have shown promising results for serial manipulators [1], humanoids [2] and also a special case of affordable personal robots, such as Pepper, used in this work, with high motor-joint backlash and friction [3]. In our previous work [4], we proposed a contact observer method that uses data-driven machine learning techniques to detect a colliding link and monitor contact direction and intensity. This method was developed to fit well into minimal sensor set-up, as only joint encoder measurements are exploited for contact detection for both moving and static links.…”

“…In this paper, we describe the modifications made to our initial method [4] in order to increase its accuracy and generalization properties in the case of multi-joint motions; we demonstrate its experimental performances in real-time feedback compliant motions regulation on the arm of a humanoid robot Pepper.…”

“…In order to address a more challenging task of contact sensing without using cover haptics, joint torque or even motors' electric current sensors; we developed a position tracking error based contact observer in [4]. This method is based on monitoring the discrepancy between the measured position tracking error and the expected one for a given desired free motion trajectory.…”