Photo Credit: Jake Peters, Orit Peleg

At the level of individual organisms, we are interested in and have studied aspects of  object recognition via shape and size discrimination,  navigational tasks in insects, the statistical nature of geometric reasoning to understand how we might have an intuitive sense of Euclid’s elements, detecting motion using the early warning signals embedded in fluctuating motion, etc.   A subject of particular interest is linking geometry, dynamics, and probability, in such humble examples as the coin toss and the ball throw, to such ethereal examples as the cognition of visual space and motion, and in minimal models of cognitive classification using invariant descriptions such as those based on statistical geometry.

We also have an increasing interest in exploring aspects of collective dynamics in functional settings, e.g. collective dynamics of cell aggregates, the behavior of super-organisms exemplified by social insects, and most recently, the design of cooperative robotic swarms. A basic question here is not so much the plethora of instabilities and patterns that they exhibit (unsurprising given the range and nature of collective interactions in sentient matter), but why they do so, and how they are regulated to achieve a modicum of functional efficiency. Patterns allow for the harnessing and control of matter, energy and information in space-time, and create micro-niches that are neither completely permeable nor completely insulated. Either extreme will not allow for function. Social insects are a great example to study these questions of homeorhesis because they occupy a range of ecological niches and are evolutionary experiments that work.  They also are not dissimilar to multicellular organisms with a division of labor, analogs of somatic and germlines, etc. And they can be studied in the lab and the field at multiple scales. We started with termite mounds in Namibia and India in 2012, and have since worked on bee clusters in our own Concord Field Station, and finally with ants and robots in the lab, always integrated within a theoretical framework.  Studying animals in their environment makes obvious the artificial nature of the arbitrary boundaries between the living and non-living – behavior changes the (micro)environment that then feeds back on the behavior … which we have measured and quantified in these different social insects, and are now attempting to synthesize in robot swarms.

Our first foray was to understand the dynamics of active porous media, where flow changes the porosity which then affects the flow itself –  inspired by wall and channel building in ants, as well as thermoregulation in bee clusters.

These theoretical studies on ant and bee behavior were followed by a series of experimental field studies on termite mounds, starting with Macrotermes michaelseni in Namibia, and Odontotermes obesus in India. Our observations showed that termite mounds breathe like lungs – driven by diurnally-driven thermal convection flows that overturn everyday, i.e. converting temporal oscillations to spatial flows. This is akin to heat engine that operates between oscillating sources and sinks, the norm in biology, with lessons for technology.   We then used simple theories to predict how the size and shape of termite mounds scale with species type, population and geography. And most recently, we have constructed continuum models of termite biotectonics to explain the appearance of ramp-like (edge-dislocations) and helical (screw dislocation) morphologies in the terraced interiors of mounds, modeling the termites, air and building mud as three interacting fluids, with the resulting statistics similar to those observed in nature.

Bee clusters have solved complex physiological problems posed by the need to regulate their temperature (thermoregulation), gas concentrations (ventilation) and stability (mechano-protection). We have studied these using experimental and theoretical approaches to understand how clusters flatten collectively (mechanical altruism?) in response to mechanical loads (reminiscent of  “Once more unto the breach, …”), how they drive active ventilation patterns by fanning their wings at hive entrances, and how they change their shape and size in response to heat loading.

Recently, we have started to study cooperative behavior and task completion in ant collectives, inspired by the ability of ants to dig out of traps by tunneling. Our study showed how ants can modulate the environment by leaving pheromone signals that then modulate their behavior. This leads to the emergence of embodied cooperation and a type of physical intelligence.  To recreate this in a minimal setting, we built a small collection of light-driven R(obotic)ants that leave behind and respond to a photormone signal (a visual analog of a pheromone) and used them to show how simple local rules for sensing and action lead to a rich phase diagram that captures the dynamics of their biological analogs. This  builds on our early attempts to study cooperative behavior in   bristlebots and  to understand what are the minimal rules that drive functional patterns that can sense, act, learn and adapt.