Monday, March 29, 2021

The Flight of the Smellicopter

An interesting article popped up today in Electronic Design's email news, about navigating micro drones by using a sense of smell. Navigating a drone or some equipment using smell has a lot of practical uses; the obvious one would be looking for a gas leak in a place with natural gas utilities, but there are others, such as the way bloodhounds have long been used for tracking people by scent. What about a missing child? What about environmental hazards, chemical leaks, and other industrial accidents?

The problem with using smell to navigate is that while we do have things that function as synthetic noses, they aren't very good.  (Where's my drummer?  They just don't smell good. (rim shot))  That's why a team led by Ph.D. candidate Melanie Anderson at the University of Washington thought they'd use the smell-sensing antennas of a Manduca sexta moth to produce a drone that they call the Smellicopter.
Why a moth antenna? In addition to sensing wind and vibrations, these fast-responding, highly sensitive transducers capture olfactory information that the insect uses to find food and mates. A sensed odor induces a complex series of chemical reactions, culminating in an “action potential” that propagates down the antenna to the brain of the insect. An electroantennogram (EAG) measures the aggregate electrical activity of the olfactory neurons in an antenna by measuring the voltage drop across the antenna.

“Nature really blows our human-made odor sensors out of the water,” said lead author Melanie Anderson, a UW doctoral student in mechanical engineering. “By using an actual moth antenna with Smellicopter, we’re able to get the best of both worlds: the sensitivity of a biological organism on a robotic platform where we can control its motion.”

That point was amplified by co-author Thomas Daniel, a UW professor of biology who co-supervises Anderson’s doctoral research. He added, “Cells in a moth antenna amplify chemical signals. The moths do it really efficiently—one scent molecule can trigger lots of cellular responses, and that’s the trick. This process is super-efficient, specific, and fast.”

If you hear of the University of Washington being attacked, I wouldn't be surprised if Ms. Anderson's lab at UW were to be set upon by the animal rights people because she's plucking the antennas off live moths for this experiment.  In this photo, the antenna is the little round loop on forward (left) end of the drone.  The moths are put into a refrigerator to anesthetize them before the antenna is plucked off, but there's only two antennae per moth so the moths are either regenerating the antenna or they're sacrificing the moths. 

All that aside the way they handle the signals from the moth is plain old analog electronics.  They measure the output of the antennas. Output was between 10 µV and 1 mV in response to stimuli, so they set the gain to 1,000 after some experimentation. The moth's antenna output impedance was high, though, between 500k and 750k ohms (500 and 750,000 ohms), and the system picked up electrical noise too easily.  That was easily solved with an active filter and more amplification - the output stage has a gain of 11. 

The drone pictured with the modification for using the moth antennae is a commercial/open source model called the Crazyflie 2 Nanodrone.  Without the antenna sensor (another EAG, or ElectroAntennoGram) it weighs 23 grams.  Once outfitted for their experiments and characterizations, it can fly up to seven minutes on its 250 mAH battery. 

It's an interesting article, digging into how the biology department told the mechanical engineers how the moths work their way upwind to the source of a smell, as well as digging into the details of how the engineers have made the drone work.  Biomimetic designs, literally life-copying designs, are becoming a big thing (as was always expected) and this a good look at how it's going on. 


  1. Once perfected such a device would be an excellent assassin's choice. It will seek out a specific odor...which could either be a persons intrinsic smell OR something planted on them.....find them and then detonate upon contact. Humans
    will ALWAYS find ways to use new technology to kill with.

  2. That is pretty damned brilliant.

  3. I'm sure I read about this in MAD Magazine back in the day . . .

  4. We're smack in the middle of my wheelhouse, amazingly enough. Sensory biology is the cure for insomnia. I discovered this by publishing a bunch of monographs in the field as a yoot.

    The Office of Naval Research gave us 5 million bucks to design a sensor package towards making mine-sniffing robots. We shied away from moths because everyone's doing moths and it's more affordable to dive into the physics of diffusion (chemical tracing) in a water medium than in air. We went underwater.
    The problem comes where we don't know how animals track odors. Chemical signal dispersal isn't just an issue of diffusion. Turbulence in the fluid column causes the signal to become patchy, creating two very distinct fields. The 'nearfield' is what you might expect, where chemical trace (signal) strength is proportional to the distance from the source- an animal can follow the path where the signal strength increases fastest over time... but turbulence causes the signal to become patchy relatively quickly, and signal strength no longer is a function of distance to the source. Instead, it is governed by chaos math. Even in our 60' long testing flume with perfectly laminar flowing water, friction on an atomic level is enough to make turbulence sufficent to disrupt linear diffusion of an odor (signal) in the water or air.
    Animals use multiple search algorithms to follow odors to their source. I discovered where lobsters make the switch between their 2 most obvious searching behavior regimes by modeling the turbulence in the water and over their mechanical-vs chemical sensory receptors. I got folks excited by grinding up HDPE in a coffee grinder, boiling it in concentrated fluorescing dye (rhodamine) then flash-freezing it in liquid nitrogen to bring its density (after thawing) to the same as 66 degree seawater. The glowing HDPE could be traced by high-speed camera in the water under a blacklight, and we modeled the turbulence... and with all that, there were at least a half dozen gross search algorithms that lobsters were using that I could recognize but not model when they successfully traced the smell of food to the source. Imagine how many I failed to recognize.

    The moth thing is still a more popular avenue for research, because it's a 3d working environment (with lobsters we had to account for friction from the floor of the tank, but a lot easier, and we could blindfold the lobsters to eliminate visual hunting and spatial memory), and moths are a lot cheaper to keep than lobsters. Anways, for 5 million bucks the ONR got a lot of half-answers, I got $5 an hour, and after a couple of years, we moved onto putting laser telemetry backpacks on sharks. Yes. Sharks with laser beams. To date, biomimeticists still can't couple chemical sensors with a behavioral algorithm reliable enough to track odors consistently. I was doing this stuff in 1997. The crap we did with MS-dos, Radio-Shack parts and an algorithm that was very crude is still as reliable as the stuff being produced today. And I'm still upset that I spent 2 summers at Woods Hole with prison pallor, in a basement under a blacklight, 6am-8pm 7 days a week.

    1. Best thing I've read in weeks. Thanks, Paul!