After a couple years of trying to pitch lofty, abstract concepts for funding from the Smithsonian Tropical Research Institute, I changed up my strategy with great success! I went from rejected proposals such as “Give me money so we can do large scale exploratory performances with crazy technology in the jungle so that people can feel what it’s like to be an ant” to “Let me build an ant sensor” and surprisingly enough it got funded 🙂
As with all things in life, I realized that the reason I was having trouble getting funded last year was due to communication problems. No one should really be expected to just flat out accept big vague ideas that they have no real connection to what’s being proposed. So after doing lots of exploratory “Digital Naturalism” stuff last year, i decided it would be a good idea to take one of the projects that developed from this process, and present a very concrete idea that could gain traction among the scientists on the review board.
The basic concept
In the lab, we can track lots of ants with cameras pointed at the colony in a nice 2-dimensional plane. In the field however, the ants live on arbitrary geometric shapes- up the bark of a round tree and onto frilly leaves, for example.
While working with Peter, we had been adopting the computer vision techniques from the lab in field sites with little success. However, after our work designing, utilizing and performing with the technology together in the jungle, we were able to start analyzing our problem from the ground up.
For Peter’s experiments, we realized that we didn’t need all the data that the lab tools were working to collect like, ant position, unique ID, orientation. Instead what we could really use would just be something that told us there mere fact that an ant was there or not. In the little time that I had last year, I made a really simple ant-detector prototype. An LED gives steady illumination to a point on the tree bark, and a photo-resistor gives a reading of how much light gets reflected back. When an ant walks in front of the area where this simple sensor is pointing, it reflects the light differently and gives a different reading.
This early prototype showed lots of promise. Once implemented, we can potentially build cheap, sub $10 sensors that could be attached in arrays to arbitrary surfaces in the jungle. This would be a different means of tracking the insect movements with its own bonuses and limitations.
|Camera Tracking||Modular Sensor Tracking|
|Single Unit – Expensive||Multiple Cheap Units|
|Rich Potential Information: Speed, Unique ID, Orientation, Multi-ant Interactions||Minimal Information from single source: Ant Present, Yes or No.|
|Single Location, 2-dimensional||Multiple Locations|
All Potential Technologies
Since I have learned the lesson over and over that everything will go wrong, and most things you assume to work will not, I came with several contingencies plans of different technologies which could also potentially work.
Building more sophisticated versions in keeping with the original LED + Photoresistor sensor.
Modulated Light / Proximity Detection
This is the next step up from the original idea. The output light is pulsed, so the sensor knows exactly when to expect readings (cutting down on noise). Depending on how these readings come back, fancy sensors like the VNCL4000 (https://www.adafruit.com/products/466) can actually give distance
Optical Mouse Sensor
Right before I left for panama, Sparkfun started selling optical mouse sensors (https://www.sparkfun.com/products/12907). These chips are SUPER CHEAP ($1), and are actually high-frame rate cameras design to detect changes of movement in their visual fields. I ordered a bunch, and will try to see if I can rig them up to monitor patches of bark or leaves for the movements of any passing ants.
Electric Field Proximity Sensing
This option could be cool, because if successful, we could potentially detect ants within the trees themselves. This type of technology uses emitted electrical fields and senses any changes in the field strength it monitors coming back. Joshua Smith tested out a lot of this technology back in the 90’s with lots of sucess with Humans (we are big conductive blobs of water). In order to detect ants, this might not work at all (they are small dry and barely conductive). But if all other methods fail, this could be a cool thing to resort to.