Teleoperation With Haptic Sensor-Aided Variable Impedance Control Based on Environment and Human Stiffness Estimation

• Published in: IEEE sensors journal Vol. 24; no. 14; pp. 22168 - 22177
• Main Authors: Wang, Zican, Xu, Xiao, Yang, Dong, Gulecyuz, Basak, Meng, Fanle, Steinbach, Eckehard
• Format: Journal Article
• Language: English
• Published: New York IEEE 15.07.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Control systems design; Controllers; Electromyography; Haptic interfaces; Haptic technology; Haptics; Impedance; Manipulators; Position measurement; Probabilistic models; Regression models; Robot control; Robot sensing systems; Robot sensors; robotics and automation; Robots; Root-mean-square errors; sensor applications; sensor data fusion; Sensors; Signal quality; Stiffness; Task analysis; Time lag
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2024.3369758

This article proposes a novel haptic-sensor-assisted variable impedance controller (VIC) for the position-force teleoperation architecture, which measures both the environment and human operating stiffness for controller design. The VIC is widely studied and used in human-in-the-loop (HITL) applications to increase control transparency and flexibility. The stiffness adaptation strategy of the traditional VIC estimates human endpoint stiffness during task demonstrations using wearable electromyography (EMG) sensors on the operator and uses the learned stiffness profile to adjust the controller. The extra sensor complicates scalability for both experimental setups and real-world applications. In this article, we improve the performance of the teleoperation with VIC, using learning from demonstration (LfD) to generate the control strategy with human stiffness. We introduce a human stiffness estimation model that incorporates data from both haptic and robot sensors. This model generates a stiffness profile using the Gaussian mixture model (GMM) combined with Gaussian mixture regression (GMR), ensuring a comprehensive and precise representation of stiffness characteristics. To stabilize the system against the energy introduced by time-delayed communication and the VIC, time-domain passivity control (TDPA)-energy reflection (ER) is adopted to passivate the control system. For validation, vertical contact experiments with both high- and low-stiffness environments are conducted with or without time delay. The results indicate that our haptic-sensor-assisted stiffness profile successfully modifies the controller stiffness to adapt to the environments and the working situation. In addition, objective metrics such as root mean square error (RMSE) and peak signal-noise ratio (PSNR) also show that the haptic-sensor-assisted VIC improves both the signal feedback quality and the user's experience compared with traditional constant stiffness teleoperation controllers.