Computer scientists at Kobe University in Japan have built a computer that draws inspiration from the swarming behavior of soldier crabs.
The computer is based on theories from the early 1980s that studies how it could be possible to build a computer out of billiard balls. Proposed by Edward Fredkin and Tommaso Toffoli, the mechanical computer was based on Newtonian dynamics and relied on the motion of billiard balls in an idealized, friction-free environment instead of electronic signals like a conventional computer.
The model was developed to investigate the relation between computation and reversible processes in physics. A channel in this computational system would carry information encoded in the form of the presence or absence of billiard balls. The information is processed through a series of gates which the balls either bump into and emerge in a predictable direction based on the ballistics of the collision or which they don't bump into and emerge with the same velocity.
[partner id="wireduk"]
Picking up where Fredkin and Toffoli left off, Yukio-Pegio Gunji and colleagues at Kobe University have essentially built a billiard ball computer using soldier crabs. In their report (.pdf), they demonstrate that "swarms of soldier crabs can crabs can implement logical gates when placed in a geometrically constrained environment."
Soldier crabs or Mictyris guinotae live in flat lagoons and form huge colonies of hundreds of thousands of individuals. When they emerge during low tides and form enormous swarms, the crabs exhibit two different behaviors. Individuals on the edge of the swarm show aggressive leadership, keeping a solid edge to the group as they move forwards (or, more likely, sideways) in unison. Those in the middle of the swarm just follow their neighbors and so move in a more dynamic way. The crabs on the edge of the swam tend to continually fold back into the body of the swarm, only to be replaced by another.
When a swarm of crabs is placed into a corridor with walls on each side, the crabs will closely follow the wall like a rolling billiard ball. This sort of behavior can be easily controlled, for example, by casting a shadow from above on the swarm to mimic the presence of crab-eating birds. The soldier crabs will move away from any shadowed areas for fear of being munched on. When two swarms of crabs -- or "crab balls" -- collide, they appear to merge and continue in a direction that is the sum of their respective velocities.
Based on these observations of crab behavior, the team built a pattern of channels that act like logic gates. They first simulated the soldier crab swarming behavior in special patterns of channels. They then created a real system of channels in their lab and unleashed groups of 40 real crabs, which were guided using the fake bird shadow.
They found that they could build a decent OR-gate using the crabs -- this was the place where one or two crab swarms are merged into a single one. However, the more complicated AND-gate required the combined swarm heading down one of three paths. This was found to be less reliable. However, the team believe that they could improve the results by creating a more crab-friendly environment.
The findings open up the possibility of creating an unconventional computing model where the zeros and 1s are represented by the absence or presence of a swarm of crabs.
You can read the fascinating study in full here.
-- Olivia Solon
