A Self-Tuning Impedance-based Interaction Planner for Robotic Haptic Exploration

TL;DR

This paper introduces a self-tuning impedance-based interaction planning method for robotic haptic exploration that adapts to environmental uncertainties using only haptic feedback, enabling compliant contact and autonomous trajectory planning.

Contribution

It proposes a novel self-tuning impedance control algorithm with exploration and bouncing strategies for improved robotic haptic interaction under uncertain conditions.

Findings

Successful autonomous trajectory planning in unknown environments

Achieved compliant interaction despite environmental uncertainties

Demonstrated scalability across multiple scenarios

Abstract

This paper presents a novel interaction planning method that exploits impedance tuning techniques in response to environmental uncertainties and unpredictable conditions using haptic information only. The proposed algorithm plans the robot's trajectory based on the haptic interaction with the environment and adapts planning strategies as needed. Two approaches are considered: Exploration and Bouncing strategies. The Exploration strategy takes the actual motion of the robot into account in planning, while the Bouncing strategy exploits the forces and the motion vector of the robot. Moreover, self-tuning impedance is performed according to the planned trajectory to ensure compliant contact and low contact forces. In order to show the performance of the proposed methodology, two experiments with a torque-controller robotic arm are carried out. The first considers a maze exploration without obstacles, whereas the second includes obstacles. The proposed method performance is analyzed and compared against previously proposed solutions in both cases. Experimental results demonstrate that: i) the robot can successfully plan its trajectory autonomously in the most feasible direction according to the interaction with the environment, and ii) a compliant interaction with an unknown environment despite the uncertainties is achieved. Finally, a scalability demonstration is carried out to show the potential of the proposed method under multiple scenarios.