Continuous critic learning for robot control in physical human-robot interaction

Wang, Chen; Li, Yanan; Ge, Shuzhi Sam; Tee, Keng Peng; Lee, Tong Heng

doi:10.1109/iccas.2013.6704029

Cited by 10 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [15] a mathematical relation between the velocity of the human-robot interaction point and the force applied by the human operator was established using impedance control for handling tasks, where an adjustable force threshold was used to enable the operator to keep authority over the robot motion. An optimal impedance adaptation was investigated in [16] for interaction control in constrained motions, which lead to an optimal realization of trajectory tracking and force regulation.…”

Section: Literature Reviewmentioning

confidence: 99%

Human–robot collaboration for safe object transportation using force feedback

Solanes

Gracia

Muñoz-Benavent

et al. 2018

Robotics and Autonomous Systems

View full text Add to dashboard Cite

This work presents an approach based on multi-task, non-conventional sliding mode control and admittance control for human-robot collaboration aimed at handling applications using force feedback. The proposed robot controller is based on three tasks with different priority levels in order to cooperatively perform the safe transportation of an object with a human operator. In particular, a high-priority task is developed using non-conventional sliding mode control to guarantee safe reference parameters imposed by the task, e.g., keeping a load at a desired orientation (to prevent spill out in the case of liquids, or to reduce undue stresses that may compromise fragile items). Moreover, a second task based on a hybrid admittance control algorithm is used for the human operator to guide the robot by means of a force sensor located at the robot tool. Finally, a third low-priority task is considered for redundant robots in order to use the remaining degrees of freedom of the robot to achieve a pre-set secondary goal (e.g., singularity avoidance, remaining close to a homing configuration for increased safety, etc.) by means of the gradient projection method. The main advantages of the proposed method are robustness and low computational cost. The applicability and effectiveness of the proposed approach is substantiated by experimental results using a redundant 7R manipulator: the Sawyer collaborative robot.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Human–robot collaboration for safe object transportation using force feedback

Solanes

Gracia

Muñoz-Benavent

et al. 2018

Robotics and Autonomous Systems

View full text Add to dashboard Cite

show abstract

“…In this study, robot anticipated the partner's intentions to adapt the motion by task learning. Optimal impedance adaptation was studied for constrained motion HRI in [6]. Continuous critic learning was utilized for interaction control then, the desired impedance was obtained to be used as an optimal realization for satisfying control objective.…”

Section: Related Work and Motivationmentioning

confidence: 99%

Neural impedance adaption for assistive human–robot interaction

2018

View full text Add to dashboard Cite

The problem of assistive human-robot interaction (HRI) with unknown impedance parameters is nontrivial and interesting. This problem becomes even more challenging if unknown reference trajectory and uncertain robot dynamics are involved. This study investigates an intelligence impedance adaption control scheme to assist human interaction with an unknown robot system. An algorithm is proposed to facilitate assistive HRI by optimizing the overall human-robot interaction performance. Neural networks (NN) and backpropagation are employed to tackle the optimization problem, based on an online adaption of impedance parameters. The tuned impedance model is integrated into the design of the neuroadaptive controller. The controller is modified by utilizing the barrier Lyapunov function technique to increase the safety, and to improve functionality of the NN during the system operation. The obtained controller can learn the robot dynamics online while coping with both the problems of trajectory-following and impedance model-following. Stability and uniform boundedness of the closed-loop system are verified through Lyapunov direct analysis. The effectiveness of the proposed control design is validated by theoretical analysis and numerical simulation.

show abstract

“…Considering human arm dynamics and arm mechanical impedance modelling, or pHRI and, in particular, admittance and impedance control, it must be noticed that an extremely rich literature exists. For example, many different control strategies have been developed to adapt or optimize the admittance/impedance filter parameters exploiting the robot redundancy [12], applying an online fast Fourier transform to the measured forces in order to detect and avoid incipient oscillations [13], using model-free continuous critic learning [14], proposing a variable impedance filter with online identification of the human arm stiffness [15], or using neural networks to learn online from data the robot and human arm models [4,16,17,18]. On the other hand, different experimental devices and protocols have been also created to investigate the characteristics of the human operator during admittance and impedance controlled tasks [19,20,21].…”

Section: Introductionmentioning

confidence: 99%