A Friction-Driven Strategy for Agile Steering Wheel Manipulation by Humanoid Robots

Abstract

Vehicle driving can substantially enhance the maneuverability of humanoid robots. Agile steering wheel manipulation requires rapid rotation in narrow spaces such as a cab, serving as the foundation for increasing driving speed, especially in an obstacle avoidance scenario. Generally, there are 3 human driving strategies, “Hand-to-Hand,” “Hand-over-Hand,” and “One-Hand.” Based on the human driving motion data, we quantitatively analyze these strategies from 3 aspects, motion range of joint combination, motion region of the shoulder, and velocity of the manipulation. Then, a friction-driven manipulation strategy using one hand is proposed utilizing the similarity between a humanoid robot and a driver (human). It effectively addresses the requirements of both a small range of motion and rapid manipulation. To prevent the deformation of the steering wheel caused by excessive force, we construct an operating force model specifically for the steering wheel. This model accurately describes the relationship between the rotation resistance and the state of the steering wheel. In addition, we propose a quadratic programming (QP)-based control framework to servo the robot to track the end-effector position and target wrench output by this model. Finally, the effectiveness of this paper is evaluated through an obstacle avoidance scenario, achieving a maximum rotation velocity of 3.14 rad/s.