With the exception of computer automation, the design of mechanical systems has not fundamentally changed since the industrial revolution. Even though our philosophy, aesthetics, and tools are far more refined and capable, we still design mechanical systems around bars, linkages, and motors. Similarly, electronic systems have largely been constrained to rigid circuits. These constraints are the central challenge being addressed by two emerging fields: soft robotics and soft electronics. In this post I’ll talk briefly about why non-rigid systems are interesting (and potentially useful), and present results from recent work that I published in Science Magazine on this topic.
Traditional robots consist of rigid bars that are separated by joints. In this framework, we place motors at the joints, and use inverse kinematics along with motion planning to control physical state of the robot. In many ways, this design framework is inspired by our observation of biological systems. Take for example humanoid robots, which use the concept of taking rigid bones (e.g., a femur and a tibia) and separating them with a joint (e.g., a knee) to cause motion via the dynamic control of the angle at that joint. The similarities obviously break down when we consider the means by which we control the joint–humans use a complex system of contractile muscles that run alongside bones and attach to joints via tendons, while our synthetic analogue replaces those muscles with a motor at each joint–but they are designed with similar functional goals in mind.
But what if we want to develop a machine that can move like an octopus? Similarly, what if we want to make a stretchable, skin-like display that conforms to the arm, or perhaps even a system that mimicks biological skin? Bars, linkages, and electric motors don’t make sense in these cases. Deformable systems require a fundamentally different approach to design, one that challenges our basic notions of how mechanical and electronic systems ought to work.
I currently work in the Organic Robotics Lab at Cornell University, where we are addressing these types of questions. Our research falls under two broad categories: soft robotics and soft electronics.
At its core, the field of soft robotics is a design philosophy in which the goal, at this point in time anyway, is to explore new capabilities that might be offered through the use of highly deformable mechanical systems–either inherently deformable (like a rubber band) or passively compliant (like origami). Although the field is still mostly an academic pursuit, some of the technology has already been commercialized. Empire Robotics and Soft Robotics Incorporated have both developed soft gripping devices for manufacturing, material handling, and food processing automation. There is also a great deal of interest in soft machines for biomedical devices (e.g., exoskeleton suits) and surgical tools (e.g., balloon catheters).
Perhaps the most exemplary ‘soft robot’ to-date is a pneumatically actuated rubber quadruped that performed the worm under a ~1cm opening using a series of undulating gaits. This is work that my doctoral advisor, Rob Shepherd, completed as a postdoc in George Whitesides’ lab back in 2011.
Above all, this is a demonstration of extremely complex motion that is achieved using no on-board sensory feedback, and a fixed sequence of simple instructions that tell the robot which leg to actuate, when to actuate it, and for how long. While traditional robots have relatively few degrees of freedom, this piece of rubber has infinite degrees of deformable freedom. The nuance, of course, is that not all of them are accessible through actuation alone–it only has five internal chambers. However, it still can undergo dynamic shape modulation that would require extremely complex design and controls with a rigid structure.
There are a host of other examples that can be found throughout the soft robotics literature. Mike Tolley and Daniela Rus provide an excellent review of the field here. Within the last year there has been several key advancements. Most recently, a group led by Rob Wood from Harvard University demonstrated a self-contained, untethered octopus-structure, named Octobot, that relies on a 3D printed vasculature system of interconnected combustion chambers in combination with hydrogen peroxide as a fuel source to produce octopus-like motion. Here is a YouTube clip that highlights the main idea.
Interestingly, Octobot has no electronic components–it’s kind of like a wind-up toy. One of the fundamental challenges in soft robotics is embedded electronics, and obviously so; it requires electronic components that can stretch and deform. Earlier this year my colleagues and I published a paper in Science on highly stretchable displays (highly stretchable electroluminescent skin for tactile sensing and optical signaling). While the paper has a bent towards soft robotics, the concept is also relevant to wearable electronics. The initial idea for this paper was conceived during a class project back in Spring 2014 when we were trying to come up with a way to make cheap stretchable displays that could be used for wearables. Our initial ideas were crap, but eventually our notions of what was and wasn’t possible, and what was and wasn’t novel, converged on the idea presented in that paper.
The concept we explore is a hyperelastic light-emitting capacitor, or HLEC, that can be overlaid onto a soft robot to provide it with the ability to dynamically change its color, and the ability to sense both its shape as well as environmental stimuli, similar to biological skins. Shown below in Figures 1 and 2 are concept drawings of this skin overlaid on an octopus tentacle, and a cross section view of the skin, respectively.
To make the display stretchable, we embedded transition metal-doped zinc sulfide phosphors (which are used for lighting in car dashboards and exits signs), into a thin elastomeric sheet and sandwiched it on both sides with a conductive hydrogel polymer. The electrodes conduct electricity ionically. We apply an alternating electric field across the dielectric layer (~25kV/cm) to produce light emission. While the luminous power of this device is relatively low in comparison to LEDs and even other stretchable display technologies, it can stretch greater than 6x its initial length in one direction, and when stretched in two directions it can expand its surface area by more than 7x. For perspective, previous stretchable displays were capable of roughly 150% strain along one direction. Figure 3 shows this device in a uniaxial tension test.
We also took preliminary steps towards adapting this system into a display. Figure 4 shows an 8x8 passive matrix that can display combinations of pixels dynamically. While this doesn’t exactly resemble a cell phone screen, the materials we use are compatible with a number of micro-fabrication techniques including photolithography, and so making higher resolution devices from these basic building blocks is very feasible.
Moving forward there are a number of interesting lines of inquiry that one could explore. New fabrication methods for higher resolution is one area that is ripe for the picking, and indeed my co-author Shuo Li is actively pursuing this. I am particularly interested in sensing. My goal is develop systems that interact with humans through deformation. This is the topic of my PhD dissertation. I am currently working on a new approach by which information is extracted from both the shape of a soft body and how it changes in time. In my next journal publication, I posit that this is a very rich medium for human-computer interaction, and as evidence provide a learning framework that enables soft devices to interpret human input and discriminate between users. I will write more about this concept as soon as the paper is published.