Harnett Lab

Three papers that inspired our group in 2016 are connected by the theme of fluid-driven robotics. Whether the fluid is a gas or a liquid, these robotic systems need a way to (1) move the fluid around, and (2) get feedback on how much it moved, because the amount of motion depends on what the robot is touching. These are very soft squishy robots.

First off is Huichan Zhao and team’s work on optical sensing in an air-driven robotic hands in the Shepherd group at Cornell, where I spent part of my sabbatical. They were finishing it up when I was there, so it was exciting to see the outcome in the very first issue of the new journal Science Robotics this December:

In this paper, the driving force comes from an air tank, and the main innovation is sensing pressure, strain and bending with an optical signal instead of an electrical one. The most practical advantage is the ability to get the signal without having to make a soft-to-hard electrical connection, which is a weak point in soft structures that go through extreme distortions. That’s where it’s going to break most times. Unlike an electric current from a resistive sensor, light can jump across an air gap. Working toward human-level touch sensing, they demonstrated the ability to sort a ripe from an unripe tomato by touch!

But what about the actuators themselves? We need new ways to shuttle fluid around in soft systems so they’re not tethered to a gas bottle. If a computer or microcontroller is going to be the brain, it’s convenient if the pump runs off an electronic driving signal (though the most biomimetic thing would be fully squishy chemical systems). The holy grail is a small, high pressure, high flow rate, efficient, low-voltage pump that can get stepped on and still work. And it would be great if nothing bad happens when the pump fluid gets in your eye.

Electroosmotic flow at conductive surfaces is our approach to portable pumping. There are hundreds of kinds of pumps, and many different species even within the electroosmotic category. Two different ideas came out of other groups in 2016.

The Martin group uses a water-based electrolyte with conical nanopores to do electroosmotically driven pumping. The surfaces are insulating, unlike our recent work, and because of the tiny dimensions at the 22 nm diameter tip, the ion concentration at the tip depends on polarity. This means they can get single-direction flow even with an AC signal. The great benefit of an AC signal is it eliminates gas bubbles from electrolysis that would stop the pump by cutting off the electric current.

Flow goes in the direction you would think, big to small end like a funnel, although intuition is a bad guide at these dimensions. The flow rates were kind of small: 4 microliters/minute per square cm of membrane.  However, this was with a sub-5 volt signal that produced an astounding 200,000 Pa of pressure thanks to the high flow resistance of the tiny pores. This is the typical pressure in a car tire (Pascals are small) but in the world of soft robotics, it’s very impressive.

The Smela group escapes the electrolysis bubble problem with chemistry. In an earlier paper, they pioneered the approach of using propylene carbonate as the working fluid instead of water-based solutions. It has the amazing advantage of producing electroosmotic flow at up to a kilovolt DC, without generating electrolysis bubbles. And it isn’t terrible if it gets on your skin (though we haven’t tried it.) They used a cellulose membrane with ~8 micron diameter pores for their electroosmotic pumping surface. With 500-600V DC, they achieved actuation of small silicone domes with forces of 45 g, similar to lightweight switches on a TV remote control. 600V may sound like a lot of voltage, but people in the soft robotics actuator world are regularly using thin dielectric elastomer actuators at 1000 to 3000 V. The pumping rates look very fast, in the range of~ 1 ml per minute per square cm of membrane, at a respectable pressure of 15000 Pa.

This paper earned a spot high on my list thanks to the innovation, but also the amount of detail on packaging and testing the actuators. The small image here shows an actuator lifting a weight!

It’s not really lasers in space, but now we have a laser feeding into the vacuum of the electron microscope. This is great for viewing how microelectromechanical (MEMS) structures respond to infrared light (915 nm). The problem with viewing our strain-engineered MEMS structures on the workbench with an optical microscope, is that they move in and out of the focal plane. We can still collect data using optical microscopy but it’s hard to show people what’s going on, since most of the device is blurry. In contrast, scanning electron microscopes have a nearly infinite depth of field, putting the entire device in focus.

I’m getting questions about MATLAB plots. Maybe it’s project time. Although I am busy learning Python and matplotlib, I won’t hate on MATLAB–here is my tutorial on plotting your data in 3D using 3 kinds of MATLAB plots.

This time of year brings questions on which software is best for all kinds of topics. Common topics for this question are 3D modelling, microcontroller programming, and electromagnetic simulation. If you’re not told what to use in a class or at work, it’s smart to pick software that lets you get moving fast, rather than aim for the software package that’s #1 in the field. Pick something where you don’t get bogged down in setting up your work environment, whether because it’s easy or because you have local support specifically for that software, so you can move on immediately to learn the things that are more universal. The basic concepts of generating the x,y, and z data for your 3D plot are similar from system to system, while the details of your license server are not.

Matplotlib is aiming for the same features as MATLAB plots, so most of what you would learn in a MATLAB class applies to it even though the syntax is a little different. The main advantage of Python/matplotlib over MATLAB is that it’s free (this can be important after you move to a new job), the disadvantage is that the documentation is less consistent. Greater consistency has a price…probably they have more meetings at the MATLAB factory.

We continue to investigate the potential of strings and fibers added to 3D printed, laser-cut and machined parts. The most basic application is soft, flexible links between parts that wouldn’t normally bend. Beyond that, conductive materials and sliding cables are discussed in this slide set from the IDETC conference.

Here is the preprint