Design and Fabrication of Peano-Hasel Actuators to Develop a Continuous Programmable Surface

Time Period : Jan 2023 - May 2023

Page Overview : Undergraduate Research Project designing and fabricating a HASEL actuator passive matrix for haptic applications under Dr. Alex Chortos at the Bionic Interfaces Prototyping lab at Purdue.

Contributors : Pranav Parigi, Umar Mian, Jue Wang, Alex Chortos

Learnings:

• Peano HASEL actuator design, fabrication and control principles.

• Polymer based fabrication methods.

• HV control circuit design.

• Soft fluidic actuator dynamics.

• Soft electrode fabrication with PDMS

• Academic writing and presentation.

1 ABSTRACT

Peano-HASEL actuators are high strain, flexible actuators capable of supporting large loads. Their use so far has been restricted to one dimensional actuation with a single electrode per actuator. This project explores the design and fabrication of a continuous programmable surface made up of one HASEL pouch with an array of electrodes to create pixels of actuation. The actuator is fabricated using BOPP sheets with Fluorinert used as the dielectric between them. The electrodes were fabricated out of a carbon black and PDMS mixture. The first sample made consists of a simple 2x2 matrix where the effects of one cell actuation was tested. It was controlled using an EMCO HV step up transformer in combination with LabView. Preliminary observations show that the 2x2 cell shows capabilities to actuate individual cells distinctly. Further research needs to be done to understand the charge retention time constant and cross- talk effects of adjacent pixels.

2 INTRODUCTION

2.1 Peano-HASEL Actuators:

Hydraulically amplified self-healing electrostatic (HASEL) actuators developed by (Acome et al. [1] are a new class of soft muscle mimetic actuators with high strain and flexibility. They consist of a BOPP film enclosing a dielectric liquid material such as Envirotemp FR3 or Fluorinert (Dielectric and BOPP must have large breakdown field). Electrodes are applied across the actuator as shown below in Fig.1[1]. Application of a large voltage (>3kV) causes the two electrodes to “zip” together relocating the liquid dielectric within the actuator and causing an actuation. Changing the shape of the electrodes affects the zipping pattern and thereby affects the actuation type. Current research is limited to combining single actuators in complex configurations for specific applications as demonstrated by (Mitchell et al. [2]).

Figure 1: Basic HASEL Actuation Mechanism [1]

Figure 2: Scorpion Robot with multiple HASELs [2]

HASEL Actuators also exhibit self-sensing properties as there is a capacitance between the the BOPP and dielectric layers. As the material deforms, the capacitance changes thereby allowing the actuator to detect its own deformation. This feature allows for closed loop control of the HASEL actuators without external sensors or circuits.

2.2 Passively Addressed Robotic Morphing Surface (PARMS)

Programmable surfaces (PSs) consist of a 2D array of actuators that can deform in the third dimension, providing the ability to create continuous 3D profiles (Wang et al. [3]). Wang et al. utilized machine learning regression models and Finite element simulations to develop both forward control and inverse control of a shape morphing ionic actuator. The surface was created using 81 pixels which were controlled by passive matrix addressing. This means that individual rows and columns were given voltage signals to actuate instead of every pixel. This method reduces the complexity, size and cost of the control system. This work allowed for the creation of very dynamic shape morphing surfaces which can be easily controlled and implemented for a wide range of applications.

Figure 3: Ionic Actuator based Programmable Surface [3]

3 METHODOLOGY

The ionic actuators used to create PARMS by Wang et al. have a few drawbacks the most important one being the limited stress and strain capabilities of the actuators. These issues aren’t found with HASEL actuators which can support very large loads with large strain percentages. Thus, creating a ML trained programmable surface with HASEL actuators allows us to develop high strain surfaces. These surfaces have applications in haptics, medical device industry, human augmentation, etc. One of most significant challenges in developing this surface is the control design. The actuators operate at very large voltages (>3kV) which requires very bulky and expensive switching devices. The following sections detail the work done in the controller design and the early tests run on the HASEL samples.

3.1 Fabrication of HASEL Actuators

The HASEL actuators used for our tests were fabricated using 18-micron Multiplastics 5020 Biaxially Oriented Polypropylene (BOPP) film. This film has a large breakdown strength and very low elasticity ensuring that the actuator performs with maximum efficiency. The pouch is sealed using a heat press as shown in Figure 4 below. We used a carbon black and PDMS based electrode instead of the hydrogel electrode. The pouch is sealed using a soldering iron after the dielectric is inserted into it. Figure 5 shows a sample of a 2x1 array made using this method.

Figure 4: Heat Sealing Process for HASEL actuators

Figure 5: 2x1 Sample made with heat sealing and carbon black electrodes.

3.2 High Voltage Control Architecture

The design of the control system is shown in Figure 6 for a sample 6x6 matrix of HASEL actuators. Each row gets the same actuation voltage, while every column gets one grounding voltage. Based on these signals the pixel is either actuated (zipped) or not. This design is heavily inspired by work done by Wang et al. for the PARMS. We have 3 distinct input levels of 3,5 and 10 kV to allow different intermediate voltages to be applied to the pixels instead of just 10kV and ground. We can also create a virtual ground using these intermediate voltage levels which can reduce crosstalk between cells. The HV inputs come from high voltage DC DC transformers that are capable of stepping up 5-12V to 10kV.

Figure 6: Overall Matrix Control Architecture

The actual switching of the voltages occurs in the HV switching module. The HASEL actuators operate at voltages ranging from 3kV to 10kV and they need to switch between different voltage states quickly to ensure that the surface maintains its state. Traditional transistors and diodes are not suitable for such processes. To combat this, we propose a design using Reed relays and low voltage digital signals to trigger the relays. The schematic shown below is the required circuit for a single row on the surface. The first relay switches on to charge the capacitor, aka apply voltage across the actuator. The second relay is used to discharge the circuit and ensure that the charge inside the actuator is zero. Ideally a double throw reed relay could accomplish the same with one digital signal, however the limited availability of such HV reed relays makes the system difficult to realize. All connections to ground have a large resistor in series for quick energy dissipation. This HV control system is a very basic attempt at switching HV signals, which results in a complex, bulky and expensive control box. This, however, allows us to control a theoretical 6x6 array of HASEL actuators. The reason we use passive control over active control is that it reduces the number of switching relays from NxN to 2N. It also allows for cleaner packaging of the electronics and connections.

Another set of relays to ground the actuators is also needed. This is another relay connecting each column to ground which can be digitally switched on or off. Thus, if row 1 is given high voltage and column 1 a ground signal, we expect to see an actuation in the very first pixel in the top left. All of the digital signals can be given through Arduino or LabView.

Figure 7: Individual Row Schematic

4 DISCUSSION

Passive control in a HASEL matrix is feasible only if a given pixel is able to maintain its voltage and deformation values over a large time period. If the cells discharge or redistribute their charge very quickly then the control design would have to change drastically. To test this phenomenon, we made a simple 2x2 matrix sample as shown below in Figure 8.

One row and column were given a positive voltage and ground respectively as shown in Figure 9. We expected an actuation only in one cell, however there was an initial actuation along the entire row with high voltage. This initial actuation of the entire row then slowly moved into only one cell being actuated. This effect is similar to the inertial regime dynamics of HASEL actuators shown by Rothemund et al. [5]. Figure 12 below shows the viscous regime (Left) and the inertial regime (Right). Our test showed that the adjacent cell behaved in a very similar manner with an initial actuation and over zipping followed by an equilibrium around the intended pixel ( Red arrow figure 9).

Figure 12: Viscous Regime zipping (Left) vs Inertial regime zipping (Right)

This test was run without any of the relay-based controls and so we could only actuate one pixel at a time. The actuation of the adjacent pixel mentioned above might imply the presence of a significant amount of crosstalk between adjacent pixels. The large voltages in use within the system might have caused zipping to take place in adjacent pixels where the voltage difference isn’t as high as the applied voltage. Creating a virtual ground by using some intermediate voltage to ground all pixels when not in use might significantly reduce crosstalk.

While the initial dynamics of the motion showed some characteristics of HASEL actuators, the goal of the test was to identify the time taken for the charge to redistribute within the system. This was tested by actuating the pixel as shown above and then disconnecting both terminals, leaving the circuit in an open circuit. If there was immediate charge redistribution, we would see the zipping release. However, we noticed that the zipping did not release for more than 10 seconds and displaced itself only after externally displaced. Since the displacements are small and dynamic due to the presence of a liquid dielectric, the exact point of release is difficult to identify. Further work to quantify this result will require using a depth camera to track the exact time where the system releases the zipping. The depth camera data will also show us how much the crosstalk actuation affects adjacent cells.

5 FUTURE WORK

The work so far has been aimed at making a working sample of a HASEL surface. The 2x2 surface shown above is an indication that shape morphing technology with HASELs has some promise and can be used to develop high strain, high stress programmable surfaces. Next steps in this project rely on the quantification of different characteristics of actuation to identify how feasibly passive addressing can work. Using depth camera data and image processing will allow us to map the actuation states for different voltages and combinations of inputs. Developing the HV control box to perform reliably with high switching frequencies is necessary for smooth and continuous control of the system.

The depth camera data combined with Machine Learning Regression models will be critical to creating the control algorithm for these actuators. Manufacturing larger samples and testing their responses to different voltages and loads will provide enough data for the ML model to give a high-quality result. Another application of the Machine learning based training might be in understanding the self-sensing properties of the actuators. Existing work has not characterized the self-sensing behavior in a complex passive matrix surface. ML and the depth camera data might allow us to understand how the self-sensing behaviors change under different conditions and how this can be used for closed loop control.

6 REFERENCES

[1] Acome, E., et al. “Hydraulically Amplified Self-Healing Electrostatic Actuators with Muscle-like Performance.” Science, vol. 359, no. 6371, 2018, pp. 61–65., https://doi.org/10.1126/science.aao6139.

[2] Mitchell, Shane K., et al. “An Easy‐to‐Implement Toolkit to Create Versatile and High‐Performance Hasel Actuators for Untethered Soft Robots.” Advanced Science, vol. 6, no. 14, 2019, p. 1900178., https://doi.org/10.1002/advs.201900178.

[3] Wang, Jue, et al. “Design of Fully Controllable and Continuous Programmable Surface Based on Machine Learning.” IEEE Robotics and Automation Letters, vol. 7, no. 1, 2022, pp. 549–556., https://doi.org/10.1109/lra.2021.3129542.

[4] Kellaris, Nicholas et al. “Peano-HASEL Actuators: Muscle-Mimetic, Electrohydraulic Transducers That Linearly Contract on Activation.” Science robotics 3.14 (2018): n. pag. Web.

[5] Rothemund, Philipp, et al. “Dynamics of Electrohydraulic Soft Actuators.” Proceedings of the National Academy of Sciences, vol. 117, no. 28, 2020, pp. 16207–16213, https://doi.org/10.1073/pnas.2006596117.