A Robotic Testing Platform for Pipelined Discovery of Resilient Soft Actuators
Authors: Ang Li, Alexander Yin, Alexander White, Sahib Sandhu, Matthew Francoeur, Victor Jimenez-Santiago, Van Remenar, Codrin Tugui, Mihai Duduta
Organizations: Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, ON, M5S 3G8 Canada · Institute of Materials Science, University of Connecticut, Storrs, CT, 06269 U.S.A. · School of Mechanical, Aerospace, and Manufacturing Engineering, University of Connecticut, Storrs, CT, 06269 U.S.A. · Material Science and Engineering, University of Connecticut, Storrs, CT, 06269 U.S.A. · Inorganic Polymers Department, Petru Poni Institute of Macromolecular Chemistry, Iasi, 700487 Romania
Short lifetime under high electrical fields hinders the widespread robotic application of linear dielectric elastomer actuators (DEAs). Systematic scanning is difficult due to time-consuming per-sample testing and the high-dimensional parameter space affecting performance. To address this, we propose an optimization pipeline enabled by a novel testing robot capable of scanning DEA lifetime. The robot integrates electro-mechanical property measurement, programmable voltage input, and multi-channel testing capacity. Using it, we scanned the lifetime of Elastosil-based linear actuators across parameters including input voltage magnitude, frequency, electrode material concentration, and electrical connection filler. The optimal parameter combinations improved operational lifetime under boundary operating conditions by up to 100% and were subsequently scaled up to achieve higher force and displacement output. The final product demonstrated resilience on a modular, scalable quadruped walking robot with payload carrying capacity (>100% of its untethered body weight, and >700% of combined actuator weight). This work is the first to introduce a self-driving lab approach into robotic actuator design.
Figures & tables
Figure 1: Linear DEA optimization pipeline. Batches of linear DEA samples are fabricated and tested using the DEA testing robot. lifetime is scanned across the parameter space defined by voltage, frequency, and material choices. The optimal parameter combination is then used to scale up the actuator in size and layer number, achieving higher displacement and force output. Finally, the optimized actuators are integrated into a quadruped walking robot.
Figure 2: The linear DEA testing robot. A, the testing robot. B, the block diagram of the system architecture. C, the top-down and side views of the testing robot. D, the photos of the top-down and side views of the testing robot.
Figure 3: Parameter optimization for optimal lifetime and mechanical output. A, layered optimization structure. B, Optimization result comparison. Lifetime and average displacement at 1Hz ( 40,V/μm ) and 50Hz ( 45,V/μm ) of best, baseline (default) and worst material combinations are presented. C, 3D model of the scaled up DEA with reinforcement ribbons. D, comparison of displacement under the same electric field for scaled-up DEAs with and without reinforcement ribbons. E, displacement and force degradation of scaled-up DEAs at 1Hz, above 40,V/μm , for 10 hours of continuous actuation.
Figure 4: The Modular quadruped walking robot. A, Robot design and assembly. Base locomotion units serve as the building blocks; two units form a double-unit module, and two double-units assemble into a quadruped robot. B, Gait mechanism of a single locomotion unit. C, Frequency response of walking speed for 1-, 2-, and 4-unit robot configurations. D, Long-term speed stability of a base locomotion unit. E, Payload-dependent walking speed of the tethered quadruped robot. F, Quadruped robot carrying onboard power electronics designed for untethered operation.
Figure 5: The vision of a generalizable device-level SDL for DEA design optimization . A combinatorial fabrication robot (under development) produces device samples, which are transferred by a dexterous handling robot to the testing platform. The high-throughput testing robot conducts comprehensive lifetime measurements, while an FPGA-based node manages data collection and uploads data to a database. Data-driven algorithms are used to generate new fabrication parameters for subsequent iterations. The process converges toward an optimized material and device design, which is then transferred for robotic application.
Figure S1 : Linear actuator assembly. The multilayer actuator is fabricated using standard procedures well-documented in literature [ 41 , 9 ] . After trimming, the multilayers are rolled and placed between two end caps. Carbon tape provides an adhesive interface for conductors, connecting stranded copper wires and conductive sheets. The wires serve as external electrical leads, while the conductive sheets extend the effective contact area with the DEA. A conductive filler layer is applied between the conductive sheet and the multilayer cross section; this layer is critical to the device’s cycle-life performance and is a key focus of material selection in this work. The actuator expands linearly along its length when an electric field is applied across its terminals.
Figure S2 : Data acquisition circuit for the testing robot. (Left) the photo of the circuit board. (Right) The circuit topology.
Figure S3 : Programmable power supply circuit for the testing robot. (Left) the photo of the circuit board. (Right) The circuit topology (cooling fans are not included).
Figure S4 : Testing robot architecture and workflow. (A) System architecture from user interface to hardware. (B) Data collection sequence for full electromechanical measurements. The robot periodically switches measurement modes to capture complete datasets; however, most data in this work were collected using simplified protocols measuring either force or displacement. (C) Workflow of Arduino 1 (top-layer control and DAQ board). Motor actions are shown in blue, and Arduino 2 (high-voltage board) actions in green.
Figure S5 : Displacement performance of the scaled-up actuator. (A) Comparison of DEA actuators with and without ribbon reinforcement at 42 V/μm . No significant difference is observed between the two configurations, both achieving about 2 mm displacement (8.7% axial strain). (B) Displacement versus frequency response of the scaled-up actuator at 45 V/μm , obtained using a sweep signal ranging from 1 Hz to 100 Hz over 100 seconds.
Figure S6 : Circuit diagram for untethered quadruped robot power electronics. The top layer of the board consists of MCU, battery voltage converter, and high voltage MOSFETs for discharge. The bottom layer of the board consists of two high voltage converters (HVM SMHV0520) and 3 high voltage opto-couplers (HVM OR-100).
Figure S7 : Capacitance–frequency sweep measurements of device samples under different electric fields and actuation frequencies. Measurements were performed using Analog Discovery 3 modules mounted on the testing robot. The left plots show capacitance before testing, and the right plots show capacitance after testing. Higher actuation voltages lead to greater capacitance degradation, while the influence of actuation frequency on degradation is less pronounced.
School of Mechanical Engineering, Hebei University of Technology, Tianjin 300401, China · State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
University of Connecticut, School of Mechanical, Aerospace, and Manufacturing Engineering, Storrs, Connecticut USA · Institute of Macromolecular Chemistry Petru Poni; Iași, Romania · University of Toronto, Department of Mechanical and Industrial Engineering, Toronto, Ontario, Canada