Robotuna, or How Do Fish Swim So Fast?

Ah, the fleet tuna, master of the open ocean.  This graceful, powerful fish is best known to Americans as an excellent rare steak, a common sandwich filling, or maybe a hazy memory of cartoon pitchmeister, Charlie the Tuna.  Yet the tuna is actually quite a fast and efficient swimmer with a yellowfin once making over 40kts after being caught on a line.

The thought has occurred to people that fish may be worthwhile of study if you're interested in better ways to propel an underwater vehicle such as an AUV.  Back in 1995, designer David Barrett built a robot fish for his PhD thesis.  This was the original RoboTuna named Charlie I and it became the subject for some interesting scientific work - and later a great article in Smithsonian magazine (link below).

The story starts as you might expect, with a good quality robotic fish being built.  This is MIT after all and their machine shops can turn out some of the best robotic fish imaginable.  If you'd ever imagine such a thing!  This one was Lycra-sheathed and its articulated body mimicked the behavior of a real tuna.

But after testing the fish in a tow tank it was found to be lacking in efficiency.  The whole point of looking at fish was to learn what nature-designed efficiencies might be appropriated.  But this clockwork fish could not possibly replicate the 20g bursts of acceleration or fast attack sub speeds of a biological tuna.

The researchers turned to using a genetic algorithm, evolutionary software, to control Robotuna.  The fish had originally been controlled in a standard manner: moving the tail as they saw fish swimming.  Now the genetic algorithms tested many different control techniques, starting with absolute gibberish that didn't work at all but generation after generation improving until finally the robotic fish was swimming with much of the efficiency of a real fish!  How did this happen?  After observing Robotuna and doing some visualization work in the towing tank it was found that the new Robotuna control algorithm was taking advantage of self-created vortices.  A flick of the tail to the left created a vortex, a swirl in the water, and the fish could feel and locate this vortex as it moved and at the right moment flick its tail back and push off from it.  This was one trick which helped explain the fast standing start speeds of some fish species.  It is also useful in steady-state swimming and provides a sizable increase in propulsion efficiency. 

Anyway, this is a brief description of the Smithsonian article contents and I should state that it has been several years since I last read it and so may be off on some of the details.  I may go pay for the article download now from Scientific American...

MIT Robotuna Website

Smithsonian magazine article...

PBS article

Scientific American article (paid download)

Photo credit: NOAA