Human–robot shared control for humanoid manipulator trajectory planning

Abstract

Purpose This paper aims to propose an intuitive shared control strategy to control a humanoid manipulator that can fully combine the advantages of humans and machines to produce a stronger intelligent form. Design/methodology/approach The working space of an operator’s arm and that of a manipulator are matched, and a genetic algorithm that limits the position of the manipulator’s elbow joint is used to find the optimal solution. Then, the mapping of the operator’s action to that of manipulators is realized. The controls of the human and robot are integrated. First, the current action of the operator is input. Second, the target object is predicted according to the maximum entropy hypothesis. Third, the joint angle of the manipulator is interpolated based on time. Finally, the confidence and weight of the current moment are calculated. Findings The modified weight adjustment method is the optimal way to adjust the weight during the task. In terms of time and accuracy, the experimental results of single target obstacle avoidance grabbing and multi-target predictive grabbing show that the shared control mode can provide full play to the advantages of humans and robots to accomplish the target task faster and more accurately than the control merely by a human or robot on its own. Originality/value A flexible and highly anthropomorphic human–robot action mapping method is proposed, which provides operator decisions in the shared control process. The shared control between human and the robot is realized, and it enhances the rapidity and intelligence, paving a new way for a novel human–robot collaboration.

A Review of Human–Machine Cooperation in the Robotics Domain

“…The use of piecewise functions is universal. Performance metrics such as hinge-loss [125], probability [66], the entropy of the probability distribution [126], and prediction uncertainty [127], [128] are used to adjust the control authority, which is used to determine the grasping target in the process of teleoperation.…”

“…The family of the exponential function is the other approach to gain a continuous and smooth α. Muelling et al [113] compute α using a sigmoid function considering the minimal control contribution of a user to enable smooth, seamless distribution of the control authority. Note that an excessively large value of the derivative of the exponential function may lead to too steep of authority switching [126].…”

“…In the execution layer, the operator controls the underwater manipulator with force feedback equipment according to the feedback vision and force information and completes the target task with the assistance of autonomous intelligence. Shared control teleoperation was initially realized in [126], as shown in Fig. 8.…”

Dynamical system based variable admittance control for physical human-robot interaction

“…The recent work by Bingol and Aydogmus (2020) develops a safe HRI software based on both vision and touch to guarantee the health of people working with industrial robots. Zhu et al (2020) propose an intuitive shared control strategy to control a humanoid manipulator. By providing operator decisions in the shared control process, they enhance the rapidity and intelligence of the manipulator, paving a new way for HRI.…”

An integrated vision-based system for efficient robot arm teleoperation

“…The general concept of presenting the robot pose by calculating the joint angles is defined as inverse kinematics (Craig, 2009). Although Denavit-Hartenberg modeling (D-H modeling) is already a mature method in describing forward robot kinematics (Zhu et al, 2020), it would face extra problems when solving the inverse kinematics, such as the existence of solutions and multiply solutions. Based on this fact, commonly used approaches focus more on using certain limb's geometric relationships to obtain the joint angles (Ju et al, 2014;Reddivari et al, 2014;Xu et al, 2019).…”