Session K13: Materials That Do Things by Themselves

Lesson from smart slime: How adaptive flow networks process information for complex behaviour

Propagating, storing and processing information is key to take smart decisions – for organisms as well as for autonomous devices. In search for the minimal units that allow for complex behaviour the slime mould Physarum polycephalum stands out by solving complex optimization problems despite its simple make-up. Physarum’s body is an interlaced network of fluid-filled tubes lacking any nervous system, in fact being a single gigantic cell. Yet, Physarum finds the shortest path through a maze. We unravel that Physarum’s complex behaviour emerges from the physics of active flows shuffling through its tubular networks. Flows transport information, information that is stored in the architecture of the network. Thus, tubular adaptation drives processing of information into complex behaviour. Taking inspiration from the mechanisms in Physarum we outline how to embed complex behaviour in active microfluidic devices and how to program human vasculature. 

    

Physics for local learning

Artificial neural networks learn by minimizing a loss function with a computer to achieve the desired result. Alternatively, many forms of neuromorphic computing use local learning rules inspired by biological learning. We adopt a different approach, focusing on far simpler networks that exploit physics to both perform the forward computation and to obtain local learning rules that replace back propagation. Our Coupled Learning framework, related to Equilibrium Propagation, can potentially be implemented in mechanical and fluidic networks. It has been realized by our collaborators in laboratory electrical networks, one using digital variable resistors and the other using transistors, paving the way for micro fabrication of VLSI realizations. I will discuss back-of-the-envelope calculations for the scaling that we expect of this learning platform compared to that of conventional artificial neural networks.   

Mechanical Intelligence: Lessons from biotic and abiotic systems

The living world is filled with examples of mechanical intelligence where information processing is deeply embedded in architecture of living materials. From the smallest to largest length scales ranging from protein machines to complex behavior of single cells all the way to a walking elephant - geometrical architecture of these dynamical systems enables a unique information processing platform that is not yet seen in human design. Deciphering the key physical principles of embodied computation in the living world would not only allow us to understand cellular machines as information processing units - but also inspire an entirely new class of design of intelligent matter. With examples from our work in understanding origins of behavior in single cells and non-neuronal multi-cellular ensembles - I will draw a few principles of mechanical intelligence. Finally, I will share our recent discovery of "mechanical circuits" that embody self-learning and show a few applications of this unique class of self-learning material.   

Embodying physical intelligence in soft robots for autonomy and intelligence

Autonomy is crucial for robotic applications in search and rescue, surveillance, and patrol missions. This is particularly true for the emerging field of soft robotics that are constructed of soft materials. One of the unresolved grand challenges is to create intelligent autonomous soft robots that can intelligently interact with and adapt to challenging and changing environments without external controls and any human intervention. To address the challenge, in this talk, I will discuss integrating mechanical intelligence with materials intelligence in liquid crystal elastomer-based soft active structures for achieving autonomous motions and even self-decision-making capabilities in soft robots. Two examples will be discussed. One is to utilizing a twisted structure to achieve thermal-actuated self-navigation through complex confined spaces such as mazes. The self-navigation is achieved via self-snapping and self-turning capabilities when interacting with an obstacle. The other is to utilizing a wavy ring structure to achieve thermal- or photothermal-actuated self-dancing motion and autonomous directional locomotion. The wavy ring structure enables self-sustained snapping for continuous flipping motion. Tuning the geometric asymmetry in the wavy structures renders autonomous directional motion with both a controlled moving direction and tunable moving speeds. We showed that the wavy ring is also capable of actively adapting the soft ring shape to self-navigating through a confined space that is much narrower than its body size. The mechanics underpinning the autonomous motion will also be discussed.   

Design and Additive Manufacturing of Robotic Metamaterials

Additive manufacturing has shown the promise of freedom of design, enabling parts customization and tailorable properties where superior structural performance can be achieved with a fraction of the weight density compared to bulk materials. However, it is difficult, currently, to combine different materials (structural, dielectric, conducting and ferroelectrics) to create a complex device with multiple functionalities that responds to multiple stimuli. Unlike biological systems where functions, including sensing, actuation, and control, are closely integrated, few materials have comparable system complexity.

In this talk, I will present a suite of new multi-material additive manufacturing processes and design methodologies to create materials with controlled and tailorable structural and functional behaviors. The structural phases consist of a network of micro-unit cells which collectively influence new mechanical behaviors (from high-strength, lightweight to toughening) not seen in their native counterparts. When combined with an electronic and functional phase, these materials turn themselves into a “robotic material” and are capable of motions with multiple degrees of freedom and amplification of strain in a prescribed direction in response to an electric field (and vice versa), and thus, programmed motions with self-sensing and feedback control.

I will present the manufacturing and synthesis of these materials, as well as their mechanics and design methods underpinning their novel behaviors.   

APS Medal for Exceptional Achievement in Research Winner: Sidney NagelNature does not favor order; biology does not favor equilibrium

Disordered and far-from-equilibrium systems challenge our understanding; they appear ubiquitously and their study transcends any single discipline.   What are questions for a physicist to ask which, if answered, would bring coherence and insight about this state of nature?  There are many aspects on which one could concentrate, but one possibility is to recognize that during the approach to thermal equilibrium there is a transient period in which many characteristic features can be found.  The ability of matter to store memories, of both its initial conditions and how it was manipulated and trained, is one aspect that I find particularly fascinating.  Such memories cannot exist if the material is perfectly ordered or if it has fully reached equilibrium; thus, in some ways, memory is characteristic of this transient behavior.   If we are lucky, it might even provide a way of distinguishing between different types of disorder and different forms of far-from-equilibrium behavior.  Memories can be stored in a myriad of different ways and so it is a rich subject with a wealth of examples.  In this talk, I will try to emphasize some ways in which memory formation can be a unifying concept.