Multifeedback Control of a Shape Memory Alloy Actuator and a Trial Application

• Published in: IEEE transactions on systems, man, and cybernetics. Systems Vol. 48; no. 7; pp. 1106 - 1119
• Main Authors: Shi, Zhenyun, Tian, Jiawen, Luo, Ruidong, Zhao, Gang, Wang, Tianmiao
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.07.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Actuators; Basis functions; Bionics; Cooling curves; Crack propagation; Detection; Feedback; Flexible structures; Flux density; Force; Model accuracy; Motion control; Neural networks; Nonlinear data fusion; Prestressing; Radial basis function; Resistance; resistance feedback system; Robots; Sensors; shape memory alloy (SMA) actuator; Shape memory alloys; Strain; super-elastic shape memory alloy (SSMA); Wires
• ISSN: 2168-2216; 2168-2232
• DOI: 10.1109/TSMC.2016.2641465

Shape memory alloy (SMA) actuators exhibit high strain, high energy density, and self-sensing ability, which are promising characteristics for application in soft robot systems. This paper proposes a new SMA actuator with high accuracy that is driven by antagonistic SMA and super-elastic SMA (SSMA) wires or pairs of SMA wires that is intended to be used in miniature flexible structures. A multifeedback control methodology is presented for the actuator that uses feedback from both the SMA and SSMA wires. The prestress was chosen to minimize the gap between the heating and cooling paths of the strain-to-resistance (S-R) curve for both antagonistic SMA and SSMA. A radial basis function neural network was used to establish the S-R curve model for both self-sensing feedback and SSMA feedback. To achieve better system accuracy and stability, a data-fusion algorithm (based on the support function) was used to combine the multifeedback data. Accurate and stable motion control is demonstrated through multistep response, sinusoidal tracking, and force-disturbance tests. The experimental results show that the combination model using all the feedback has superior accuracy compared to the models that use self-sensing feedback or SSMA feedback. In addition, the maximum standard deviation within the SMA-SSMA antagonistic platform and the paired-SMA antagonistic platform can be reduced to 0.036% and 0.026%, respectively. A soft joint with two degrees of freedom is illustrated to demonstrate how the current research concept could potentially be implemented in bionic soft robots.