Introduction: The Creation of the Tortoise Tracker

Before my current tortoise, I had owned one other, Izzy. This first tortoise baffled my girlfriend and me. It seemed happy and healthy but refused to eat. We would surround him with rings of various foods recommended to us by the pet store owner, yet the closest he ever came to eating was walking over to a tomato and sitting on it. Each day that we swapped out his old uneaten lettuce for a fresh leaf, our anxiety grew until the day came when Izzy just stopped moving (of course Izzy had managed to make it one day past the pet shop’s one-month return policy, despite never having eaten anything). Though our current tortoise, Vanya, has proved herself much heartier as she nears her sixth birthday, the anxiety returns when we try to understand why she has not eaten in a several days.

Unlike a pet dog, the tortoise’s modes of inter-species communication are extremely limited. You can’t call her name, because her ears do not seem to respond to a human voice; her hard shell and expressionless face prevent any sort of body language; she makes no sounds of any sort; and her only real response to touch is to recoil into her shell. In fact, the only clue to her inner workings comes from observing her travels around our apartment. Though, as Hall mentions in the essay, Critical Visualization, “maps and visualization tools offer a means to filter and make sense of …[the] deluge of data [in which] we live,” I decided to harness the power of these tools in an effort to expand upon the scarcity of information left by this creature’s only mode of communication, movement (122). The goal, then, is to map her movement, visualized in several manners, in order to understand the tortoise’s motivations.

Aside from simply “help[ing] us to comprehend huge amounts of data,” (122) Hall also describes other significant advantages of maps and visualizations. They “allow us to perceive emergent properties we might not have anticipated…facilitate our understanding of large-scale and small-scale features…and help us form hypotheses” (122). To be able to harness these properties, I had to decide precisely what to visualize. The tortoise’s movement is the focus of this mapping project, but there are various facets of her perambulations which could be considered. For example, I wished to analyze Vanya’s voyages throughout the entirety of the apartment, but I also felt that capturing smaller details of her movements would be imperative to a full understanding of the tortoise. I also was not sure of what data I needed to collect to reach my goal of enhanced tortoise understanding. For guidance, structuring the experiments and representing the collected data, I adapted Kevin Lynch’s “The City Image and its Elements” from the world of the city to the realm of Vanya and my apartment.

Lynch breaks down any depiction of a city into five functional elements: paths, edges, nodes, districts and landmarks. Paths are potential routes of movement, edges close regions off from one another, and nodes are “the intensive foci to and from which [one] is traveling” (Lynch 47). Districts are groups of paths, edges, and nodes, and landmarks are simply points about which an observer can orient him or herself. My eventual visualizations would need to depict the tortoise’s paths against all possible paths in order to unveil the hidden nodes which are attracting or repelling her. The district and landmark elements of the produced maps would respectively group data points and orient the audience. Finally, to remedy the issue of scale, I utilized multiple levels of experimentation and visualization.

The dimensionally largest of the experiments/visualizations charts the tortoise’s meanderings as animated paths over a bird’s-eye view of my entire apartment. Though bird’s-eye views are helpful in abstracting a large area, one must keep in mind that they represent a false view of the world. For example, the Peters projection controversy, in which global maps such as the Mercator’s were attacked on the basis of “distort[ing] the picture of the world to the advantage of the colonial masters of the time” (Crampton & Krygier). My map must also be viewed critically, as it imparts an impossible view from which conclusions are meant to be drawn. By pairing this bird’s-eye visualization with the other, increasingly objective visualizations, I hope to attenuate the effects of its misrepresentations. Also to aid in the final analysis of the paths, several possible stimuli, such as heat, were measured around the apartment and can be visually laid over several of the maps. Thus, if any of these stimuli pique the interest of the tortoise, they will visually appear as a set of nodes between which the tortoise moves.

Inspiration for the next mapping, a surveillance-style display of the apartment’s actual rooms overlaid with animated paths, was drawn from the long exposure Roomba pictures, MIT’s Senseable city lab projects, and the tangled paths of taxis in the Cabspotting project (http://cabspotting.org/projects/intransit/intransit.html). By combining the animated data with real, human views of the apartment, I hoped to ameliorate the distortions of the bird’s-eye map.

The set of short term experiments highlights the filtering quality of maps, which Hall discusses in Critical Visualization. Though there are myriad factors compelling the tortoise when she is first released, this map allows the viewer to focus on a single specific stimulus and observe its effects on a constrained time and space.

The views from the tortoise-mounted cameras provide the closest study of Vanya’s movements throughout the apartment. Theoretical inspiration was drawn from 16thandmission’s representation of larger spaces through intimate portraits of specific spots, and technical inspiration came from Sam Easterson’s Museum of Animal Perspectives (http://sameasterson.com/map).

To keep the viewer oriented between these many views of the apartment, each representation features a key landmark, the blue carpet. This carpet serves as the starting point for the paths of every visualization and is prominently featured or highlighted to keep the audience correctly positioned in their mental map of the entire space.

Historically, these abilities of mapping and visualizing were reserved for the authoritative powers of nations and businesses, not the common man and his tortoise. Fortunately for me, I live in a time of great upheaval in the cartographic world. Due to technological advances, Crampton & Krygier note that, “In the last few years cartography has been slipping from the control of the powerful elites that have exercised dominance over it for several hundred years.” However, even well-intentioned amateur cartographers produce undesired results. Problems arise because maps not only make sense of the world, but shape it as well. In describing how maps have the power to make reality as well as depict it, Crampton and Krygier quote theorist John Pickles: “Instead of focusing on how we can map the subject…[we could] focus on the ways in which mapping and the cartographic gaze have coded subjects and produced identities” (Crampton and Krygier 15). Maps create a delicate relationship between the understanding and taming of the subject. Therefore, I must be careful in how I use my maps lest I slip from better understanding the tortoise to accidently forcing her into undesired situations. For instance, a correlation between heat and her paths might not mean that she prefers to be in a warmer environment; instead she could just be avoiding the effects of a different stimulus. Overall, since these maps focused more on expanding the amount of information known in regards to the subject, rather than just filtering a larger dataset, I feel that my visualizations will produce greater insight to the mind of the tortoise.

References

(throughout entire documentation blog)

"Critical Visualization", Hall

“An Introduction to Critical Cartography”, Crampton & Krygier,

http://www.acme-journal.org/vol4/JWCJK.pdf

“The Image of the City”, Lynch

http://books.google.com/books?id=_phRPWsSpAgC&dq=“The+Image+of+the+City”,+Lynch&printsec=frontcover&source=bn&hl=en&ei=kLEeS4-oKIyXtgeCl6GhCg&sa=X&oi=book_result&ct=result&resnum=12&ved=0CCUQ6AEwCw#v=onepage&q=&f=false

This American Life, Episode 110 “Mapping”

http://www.thisamericanlife.org/Radio_Episode.aspx?episode=110

Roomba Mappings

http://www.flickr.com/photos/paulchavady/3458136141/in/pool-roomba/

Museum of Animal Perspectives

http://sameasterson.com/map

Stimuli Measurment and Mapping

In order to maintain consistency throughout the experiment the readings of stimuli and tracking of the tortoise was generally done between 7 and 10pm. The reasoning is that throughout the day many of the stimuli which i am testing for will fluctuate (the heat drops at night and ambient light is quite different), so since the testing will take place over many, many days a need for consistency of environment is very important. On that note, the tortoise-cam videos were actually all taken on saturday mornings because that is when the lighting in the apartment is brightest, and best for use with the tiny camera.

Apartment

A core concept of this mapping project was simply to see how the tortoise reacted just to objects and places. A top-down view of the apartment would also be handy in orienting viewers in terms of where the tortoise is in relation to other stuff in the apartment. After I measured each of the rooms, i constructed a floor plan in illustrator from which to situate my further measurements.

Then, to catalog and fill the contents of these frames, I created a gigantic panorama-like image of my apartment as seen from the top. This required taking about 500 photographs of the floor while standing on a step ladder.

The images were then roughly pieced together using photoshop's auto-aligning and auto-blending features, but I still had to do a bit of fine-tuning.

The resulting image was around 2 gigs and had a super high resolution which i had to scale back to make a version for the web.

Heat

Recording

I began the more technical mapping process by measuring a basic stimulus of life, heat. Heat seems to be a quite basic motivator, in general things tend to die if the temperature goes too far one way or the other, and since she is a cold-blooded creature, after all, I figured that heat must play a major part of her life.

Measuring the average heat of the floor was the easiest stimulus for which to test. I simply used a thermometer (technically an alarm clock with a built in thermometer) with units to 1/10 of a degree. I would first place the thermometer in a specific location on the floor of the apartment for 10 minutes (long enough for the effects of the previously measured spot to have worn off of the thermometer). After recording the value onto a spreadsheet, i then moved the thermometer to test the next spot, until a grid of values tied to locations was obtained for each room in the apartment. The resolution of the grid (the distance between single measurements) was 3 square feet.

Visualizing the Data

So that people would be able to quickly take in all this information which i had just recorded, and make judgments concerning the movement of the tortoise, i decided to visually represent the data as graded color values mapped to their location in the apartment.

First i normalized my values so that they generally ranged from 0 at the minimum temperature (67 degrees F) to 100 at the maximum floor temperature (74 degrees F). Then, after I had already mapped the layout and panoramic floor view of the apartment in Photoshop and Illustrator, I created a new gradient mesh on top of this floorplan. Next i set the Color Picker menu in Illustrator to HSB mode (Hue, Saturation, Brightness) and set the hue to 0, and brightness to 100 (for appearing hot and red). The values from the grid were then mapped to their corresponding location on the gradient mesh with a variation in the saturation of the color relating to its normalized temperature value. Thus the colder areas would appear white while warmer zones would have higher saturation values and come off as redder and hotter. Finally, since temperature gradients are typically non-discrete, i applied a Gaussian Blur to the gradient mesh to smooth the values from one test location to the next.

Sound

Whether the tortoise was even capable of hearing was a question that often came up about the pet. She doesn't seem to be too effected by speaking to her or loud noises, but she does seem to get more active when music (especially guitar music) is being played. In my research for this project i discovered that while tortoise's do not have true ears they are capable of sensing a a range of frequencies condensed to the lower to mid-range of hearing (approximately 70-300 Hz). I also later realized that this is about the range of a normally tuned guitar, which is the reasoning behind a guitar's use in later experiments. Since she is also such a creature of the ground, i figured that any low frequencies would effect her quite greatly. The mapping episode of "This American Life" is responsible for partial inspiration in how to analyze ambient sound.

Recording

To analyze the aural palette of my apartment's floor, I took a microphone attached to a camera, and placed it on the floor underneath a large, white bowl.

I figured that the bowls wide, circular mouth would simulate the effect of the tortoise's shell in how it amplifies the vibrations of the actual floor itself. I would take a 10 second audio sample, and then move the setup to the next testing location in the same grid-like pattern as with the temperature measurements.

Visualizing the Data

Before i could represent the sound data, i had to analyze it in Adobe Audition.

Using the analysis tools available in the program I was able to retreive the average frequency and RMS power for the 10 second sample from each location.Once again, i then created a new gradient mesh in Illustrator and transposed measured values from each location to color values on the mesh. This time there were two variables that needed to be represented simultaneously, frequency and power, so in HSB mode i set the hue value to the normalized frequency, and let the power correspond to the saturation. A problem that arises in dealing with sound, however, is that sounds are typically measured on a logarhithmic scale as opposed to a linear one, so i couldn't just normalize these values like the heats by just interpolating the values against a new max and min, first i would have to linearize them. Therefore to go from the logarithmic units of decibels to the linear measurement of power in Watts i had to invert the formula for decibels:

to

and then normalize the values.After the mesh was replete with values it was given the same Gaussian blur treatment to smooth out the values of the overall gradient.

Light

Recording

Unlike the process used in measuring, analyzing, and displaying the other stimuli, i was able to make use of my measuring equipment to skip a few steps. If you recall, while i was measuring the audio for the sound map i had a microphone being placed under a white bowl. To record the audio the microphone was hooked up to a camera. While the audio from the mic was being taken in, i locked the aperature settings of the camera and aimed this at the white bowl. Since all the recorded values are eventually being transposed to either Hue, Saturation, or Brightness values of the color on the gradient mesh, i could just pull an averaged color from a digital image of the white bowl and translate this directly to the mesh.

Visualizing the data

I loaded screen grabs from each clip into Adobe Illustrator, set the eyedropper/color sampling tool to its largest value, and copied the colors directly from the image of the white bowl to the gradient mesh. Then i desaturated the entire mesh so that in the final grayscale image, ambient light is simply a feature of the resulting brightness.

Smell

Volatile Organic Compounds (VOCs) are carbon based susbtances which degrade rapidly enough to cause bits of them to vaporize and float freely in the air. In other words these are substances that are generally smelly. Since pollutants tend to decay and vaporize at a much higher rate than other matierials, people typically refer to VOCs in the context of hazardous chemicals and air pollution, but VOCs can also serve as a rough indication of the relative smelliness of an area. Since most foods give off some odor and to different extents, the reading on a VOC detector near food tends to be relatively larger than barren zones, just not quite to the extent as an industrial chemical spill would have.

When Vanya cruises the apartment she will often stop to sniff a certain object, so I wanted to test the theory that perhaps certain smells would be drawing her towards places.

Recording and Data Visualiztion

The procedure for collecting and visualizing the data for smell was identical to that of the temperature measurements except that a VOC detector in a "digital canary" device was used for the testing and the data was mapped to saturation values with overall hue set to 115 degrees (the gross smelly green color).

Electro-magnetic flux

This may seem to be one of the most supernatural of the experimental stimuli and is based soley in the fact that the tortoise really seems to like wires.

Through the history of our pet ownership, this tortoise has always seemed to want to snuggle up with a good tangle of wires, or to stomp mightily around a surge protector. One time after escaping our outdoor garden, and she was feared lost after a couple of days missing, we found her near the tangle of wires under the transformer of the apartment building next door.

Recording and Data Visualiztion

The procedure for collecting and visualizing the data for smell was identical to that of the temperature measurements except that an EMF detector was used for the testing and the data was mapped to saturation values with overall hue set to 180 degrees (the electric blue).

Also a majority of the values fell within a single power of 10 but a couple of important outliers registered much higher (for instance the zone near my computer read 60 times more flux than anywhere else in my apartment). To remedy this, I assigned a multiplicative property to the hue values of the mesh to solve this problem. Therefore, whereas most regions scored on the hue value of 180 degrees (blue) the area near my computer was tweaked to a hue of 240 degrees (Magneto purple).

Constructing the Room Paths

While the top-down view of the tortoise's progress presents a clear picture of her trails on overlaid stimuli maps, I also sought to visualize an averaging of her movement in a familiar and concrete fashion. I originally got the idea of representing her progress in typical human perspective views of rooms from Flickr user Bartlec who used long exposure photography to capture the paths left by his roomba as it cleaned his house.

The problem with recreating this effect with the tortoise however, is that she is just far too slow to create as dynamic of a visualization as the roomba. Instead of a complex web of fibers visually representing the roombas internal algorithms, the tortoise's long exposure shot would simply look like a single bright line against a dark background. Therefore, instead of focusing on a single trip of the tortoise, i would establish an array of video cameras to track many of her journey's and combine these results into a more engaging piece.

A harness with two LEDs was constructed to aid the eventual motion tracking algorithms. By using a pair of lights it would be possible to not only track her location on the screen but also her relative scale and rotation.

When the cameras were ready, i would release the tortoise and wait until she had stopped in a single location for more than 10 minutes. After fifteen paths had been recorded over several nights, the video was brought into Adobe After Effects. I grouped the recorded trials by room location and analyzed the location of the tortoise in each frame by using After Effects's built in Motion Tracker. This left each clip with a custom motion path (group of keyframes representing a specific location and time) to which i could attach a particle generator. The final effect is that wherever the tortoise moved in the room, a bright path will be overlain with the thickness of the path demonstrating the tortoise's current speed (thinner lines=faster motion).

These paths were then exported as FLV files and incorporated into the final flash piece and are able to show both still images of all the paths and also animations of the paths propogating through each room.

Initial Movements

The previous experiments, by mapping the tortoise's movements across the entirety of the apartment, give a view into the long terms effects of stimuli upon the tortoise. In monitoring her movements, though, one sometimes gets the feeling that perhaps there is no master plan or memory involved, and she is just being drawn from one stimulus to the next. In reality there is probably a combination of being immediately compelled by certain stimuli and seeking out a broader goal, but I felt it would be interesting in the project to study her very first movements upon being released.

In general, I tend to set the tortoise down to begin her journey on a big blue rug in the middle of the living room. Fortunately, this carpet has a pattern of stars on it which form a grid. This is perfect for an analysis of her movements! The idea is to see what motivates the tortoise in the short run by placing representatives of the experimental stimuli equidistantly around her. With five stimuli total i created a pentagon on the carpet, the center of which would be her starting point.

Now before we could being testing her against the additional stimuli, we first had to calibrate the experiment for whatever stimuli were already present. For this reason the full experiment has a control group of paths for which the only stimuli present were the background heats, sounds, smells, lights, and electro-magnetism already recorded.

Each condition, with and without additional stimuli, featured 10 trials, and her intial orientation was shifted in each trial to ward off any biases dependent on which way she was pointing. The tortoise was set in the center and allowed to wander freely until she had stopped for 5 minutes, 10 minutes had elapsed, or she had traveled out of the testing area (the carpet).



In the same fashion as the paths experiment, where her progress was tracker through each room of the house by an array of cameras, the tortoise's motion paths were captured using the motion tracking tools in Adobe After Effects.

The resulting motion paths were then attached to a particle generator in After Effects, creating animated paths that follow the tortoise's trail.


As in the experiment with the top-down view of the apartment, gradient meshes were once again constructed in Adobe Illustrator to visually represent the stimuli that Vanya may be experiencing. These specific meshes were created using the previously recorded data (for the condition without stimuli), and data specific to the pentagon of stimuli.




Using the overlay of the different stimuli, one can begin to analyze possible attractors and repulsors concerning her initial movement.


Note: I used a guitar as the representative for sound upon discovering that the frequency range of a guitar is quite similar to the frequency range able to be picked up by the tortoise's internal ears. Since the guitar is outputting multiple, superimposed frequencies the visualization displays the average frequency output (around 175 Hz) for the guitar.

Third Tortoise POV's

The things that I felt were most missing from the "first tortoise" videos (with the camera mounted directly onto Vanya's shell) were the subtle movements she performs when encountering objects in her environment. If I was going to really try to figure out what her interests and understandings are, I needed to figure out a way to closely monitor her every movement as she maneuvers through the apartment.

Therefore I created a special harness with offset camera that would not obstruct her movement while permitting the closest view obtainable of her actions.

First Tortoise POV's

The tortoise cam videos represent the continued collapsing of the spatial scope over the course of the experiments. I already analyzed her paths over the entire apartment and her decision making process in my "pentagon of stimuli", now i sought to get even closer and even embody the subject. I feel that one of the inherent flaws of the other experiments is that they still show the tortoise reacting to our world. Perhaps to best understand the creature it would be necessary to visualize the space as the tortoise reacting to her world. Instead of labeling rooms like, "kitchen" for instance, perhaps i should have taken a more tortoise centered approach and referred to it as "big white smelly, warm room."

I desired a way to represent world from the tortoise's perspective so that maybe, when examining the environment from this novel point of view new truths would be uncovered revealing her inner motivations.

One way to accomplish, this task was to create a set of cameras which would allow a viewer to intimately follow along with the tortoise on her voyage around the apartment. One of the cameras would be offset and directed towards Vanya to permit the monitoring of her every movement. The other would be mounted directly onto her shell to give the experience of almost looking through the tortoise's eyes.


The "first tortoise" perspective was achieved by mounting a bullet camera and tiny microphone atop a soft velcro harness on the tortoise's shell. This tiny camera fed video through a lightweight wire to my larger camera which acted as a recording device.

Ideally these videos would have been created using a wireless camera so that i would not have to follow her around, but i was able to stay about 5-6 feet away from her, and I do not think she noticed. Also it would have been good to be able to mount a camera directly to her head so that you would be able to see whatever she was actually looking at, but cameras aren't quite small enough to not horribly encumber the tortoise.