Audience View Tool - Pecha Kucha at Confluence Park

This is a Grasshopper tool for determining the field of view in unconventional performance spaces. For example, Confluence Park is a new pavilion at the confluence of the San Antonio River and San Pedro Creek. Designed by Lake Flato Architects and MATSYS Design, the main pavilion is composed of over 20 huge tilt-up concrete petals that funnel water into an underground basin. After the park's grand opening in March, it will host a Pecha Kucha talk under the pavilion. We will need to seat 500 guests in a space with rather unusual sight lines and acoustics.

Photo by Sunnie Diaz

Photo by San Antonio River Foundation

Step 1: Take a section through the pavilion at seated or standing height.

Step 2: Draw the location of a screen or podium.

Step 3: The isovist component, with a test point placed directly in front of the middle of the screen or podium, determines all possible lines of sight.

Step 4: The angle of incidence on the screen is limited to 60 degrees in either direction.

Step 5: All isovist points are connected with a polyline to determine the available area.

Step 6: Add multiple screens or podiums to see the final result.

Step 7: Calculate available area with views to either screen.

The script is available here.

Try and break it.

Biometric Mapping II

This is Part 2 of a series on my Biometric Mapping project. See the first post for an introduction.

Biometric Mapping is about selecting one or more metrics, such as temperature, light, heart rate, or neural activity, and mapping it with some kind of positioning system. These can be mixed and matched to produce devices for mapping the unseen characteristics of different spaces. So far, I've adopted Arduino as the platform for collecting these metrics. Arduinos are single-board micro controllers that can host sensors and run software to record the data they collect. 

For the first experiment in mapping, I paired temperature as my metric with time-lapse photography as my positioning device. I have an Arduino board with a temperature sensor that illuminates one to three LEDs based on the air temperature in that location. By attaching a 9v battery as a mobile power source, I can then set up a long exposure photo and carry the Arduino around the room for 30 seconds. The warmest places in the room activate the red LED, and the coolest places in the room activate the green. Success! Now on to a more ambitious pairing.

Next, I want to measure temperature and daylight with GPS as my positioning system. Unlike the LED long exposure photography, I will actually have to record the data I collect and sync it to a time stamp and GPS coordinates.

This is how the Arduino is set up. I have a battery pack for mobile power (6 AA batteries are more efficient than a 9v), a TMP36 temperature sensor, and a photocell for detecting light. I need a real-time clock for recording the time, and I need a way to store all of this data. I'm using an Adafruit data logger shield to keep time and record it to an SD card.

The data logger shield requires some header pins to be soldered onto it so it can be inserted on top of the Arduino. I edited some example Arduino programs so that it would record light, temperature, and a time stamp once a second, and uploaded the program to the Arduino. Once the code is uploaded, it stays on the Arduino and immediately begins working when the battery pack is plugged in. There are GPS shields for Arduino that would allow me to record location as well, but for now, I'm using a GPS app on my phone with a time stamp, and running the two simultaneously. Now it's ready for data collection!

The setup is a little conspicuous, so I found it to be more discrete and efficient to set it up on the front of my car or bike, with a paper shield to hide all the fun bits, leaving only the temperature and light sensors exposed. Then I go for a joyride through whatever space I want to map.

Now for the second half of the workflow. This part came a lot more naturally to me, because I spend much more of my time with Excel and Grasshopper than I do with Arduino. I upload the GPS path from my phone to the computer, and upload the data from the SD card. Both of these come in the form of a .CSV, which I combine into one file.

I read the .CSV file through Grasshopper, and break out each data point: Time, Light, Temperature, Latitude, and Longitude. This particular joyride was 30 minutes, or about 1,800 seconds, which means 1,800 data points. If this becomes too dense for some operations down the line, I add a sift pattern to only take every other data point, or every third data point, and so on. I'm also choosing the domain of each, because my phone GPS and Arduino don't start at exactly the same time, and I'll need to calibrate it slightly to sync. I'm converting degrees latitude and longitude to feet, because working with true degrees is awful in Rhino. A big issue I had with this setup is that the GPS app on my phone didn't always record a data point every second. It was more like two out of every three second. So I had to interpolate that data and re-plot it to get a single data point every second. Finally, I'm plotting the location with a value to get paths like this:

Light in the Pearl District, San Antonio

Temperature in the Pearl District, San Antonio

Once I have all of this mapped, I'm using Proving Ground's Non-Linear Regression component, a fun new tool from LunchBoxML. Nonlinear regression takes several three-dimensional points (X=longitude, Y=latitude, Z=light or temperature) and interpolates the rest of the surface.

Nonlinear regression to interpolate the temperature of the rest of the space, as the domain expands.

Nonlinear regression worked very well for this, it just just takes some fiddling with the sigma value to get meaningful results. A low sigma value (left) is very local and does not smooth itself out or interpolate very far from a real data point. A high value (right) smooths things out a little too much for this scale.

By overlaying the light and temperature maps onto a plan, I get a somewhat accurate map.

Temperature at the Pearl

Daylight at the Pearl

I consider this a good proof of concept, and now I'll have to think carefully about what tests would be most meaningful to do next. These maps don't yet tell me anything surprising: It appears to be darker under bridges and behind tall buildings. This is about the scalar limit of GPS. Being in or near a building seems to interfere with its accuracy, so if I want to actually map interior spaces with more precision, I will need to explore other positioning options besides GPS.

The more passes I take in a location, the more accurate the results will be. This is not a bad amount of data for a half hour bike ride in a space this big, but it would be interesting to get more accurate and nuanced results in a smaller space.

Next steps are too think about what other metrics to measure (perhaps incorporate some biometrics?) and to consider alternative positioning systems.

Fixing Co-Planar Surfaces in DIVA: Solar Radiation

DIVA 4 for Grasshopper performs daylight simulations and solar radiation simulations for the buildings we design. It's surprisingly fast, powerful, and easy to use. But there's one annoying bug when testing for solar radiation on massing options. The DIVA solar radiation component requires that all surfaces in the Rhino model are facing the correct direction. Even when all faces are correctly oriented, the test grid is identical to the surfaces being tested, which occasionally results in dark stripes when some parts of the grid are barely on the interior of the envelope.

To fix the first issue, all surfaces in the Rhino model must be facing towards the exterior. The surfaces will appear slightly darker in Rhino if they are inside out. Notice the subtle differences in a few of the surfaces between the models below. If the color isn't apparent, use the "Dir" command in Rhino to view the directionality of each surface.

For the coplanar issue, we have to fix this in Grasshopper. Below is the typical script I use for solar radiation. I'm using dynamic pipelines so I can use a slider to add the correct prefix on each layer to load the massing option I want. For example, floors on option 1 are called "1 Floors" in Rhino. Walls on option 2 are called "2 Walls" and so on. This way, I can set the materials once, and change the prefix with a slider to move to different options.

All of those objects with materials assigned go into the Object input in Radiation Map, and as outputs, I'm calculating total floor area, total envelope area, and total radiation, as well as producing the colored grid.

The fix comes right before creating a test grid with the envelope geometry. First, I make a sphere that surrounds the entire building. Next, I find the centroid of every surface of the envelope, and find the vector between that point and the closest point on the sphere. Next, I test to see if any of these vectors are colliding with another surface on the envelope. In other words, I test to see if each surface has a clear path towards the outside of the model, or if it happens to be an inward facing surface. 

I move all outward facing surfaces just slightly along their vector, and I move all inward facing surfaces just slightly opposite of their  vector. This inflates the model so every grid surface is slightly offset from its original test surface so that there is nothing coplanar between them.

A successful solar radiation test:

The script is available here: DIVA Radiation Surface Offset.gh

Of course, you will need DIVA installed on your computer for those components to work, but if the corrective loop is useful for something else, it can be taken independently. Please try to break it and tell me how it works.

Climate Mapping with Machine Learning

Machine learning is being used to classify DNA sequences, recognize speech, and perform medical diagnoses. So far, all I can do with it is tell you how likely it is to rain in your corner of the country this year.

in 2017, Proving Ground released a new Machine Learning addition of LunchBox, their Grasshopper plugin, called LunchBoxML. This includes a few open source components built on the machine learning library Accord.NET. One of the simplest and most useful tools is the non-linear regression algorithm. 

Non-Linear Regression, Proving Ground.

Simply put, it uses a few training points to predict what the rest of the surface might look like. This lends itself perfectly to a problem I've worked on: Taking data from a few weather stations across the US to predict climate data for every site in between. There are two problems with EPW weather files. The first is that they are only available for relatively few locations, sometimes hundreds of miles apart. It's usually acceptable to use the nearest weather station, but it would be nice to have unique data for all of our building sites outside of major cities. The second problem is that EPW weather files don't include rainfall data. To make the automation of climate analysis possible, the best solution would be to enter the latitude and longitude of a site, and then let the algorithm predict the weather data based on an EPW library, and the rainfall data based on another rainfall library.

I started with a relatively small rainfall library. I have monthly rainfall averages for 26 different cities in an Excel spreadsheet, each of which have a longitude and latitude.

For these visualizations, I'm looking at the annual total rainfall, from 8 inches in Phoenix, to 67 inches in Pensacola.

The Grasshopper script reads the Excel library (on the left) and gets plugged into the non-linear regression algorithm (on the right). The rest of the components are for displaying the tags and results.

Now it's time to fine tune the Sigma value, the degree of the prediction curve. This is one of the inputs into the regression function. A low Sigma value will cause the effect of each data point to "decay" too quickly and not affect it's surrounding test points.

A high Sigma value decays too slowly and smooths out the test surface too much.

Somewhere in the middle, it starts to look like a reasonable model.

This model will benefit a lot from more data points. Increasing the library from 26 cities to a few hundred will make a much finer resolution model. This is especially important for the western US. Whereas the eastern US is much more predictable, the west has more extreme topography that creates more locally unique micro-climates. Now, it's a simple matter of adding more data points to the library, and it will begin to look much more like the map below.

Spatial Climate Analysis Service, Oregon State University

Quality Views 1.0

This is a new iteration of the quality views tool. It's more complex than the previous versions, which only used the Isovist component in Grasshopper, but it's more computationally efficient for large plans with a large number of walls, windows, and test points.

Download Quality Views 1.0.3dm

Download Quality Views 1.0.gh

Kinetic Bus Shelter

Park(ing) Day is an annual worldwide event in which artists, designers, and citizens transform metered parking spaces into temporary urban parks for a day. The activity is designed to call attention to the need for more urban open space, and to generate debate about how public space is allocated, managed, and used. This year, Lake Flato designed and built a kinetic bus shelter prototype as one of the seven pop-up parks in front of 1221 Broadway.

Throughout most of the day, the shade structures on existing VIA bus stop shelters don’t actually shade the bench below it. Because of the daily and annual change in the sun’s position, the bench is only shaded for a few hours a day during the summer.  For Park(ing) Day, our team designed and built a kinetic bus shelter to address these issues and to serve as a prototype for a VIA bus shelter to be installed as part of the Dignowity Park Project by Public Space East.

A dynamic bus shelter provides a more comfortable waiting environment, which leads to greater rider satisfaction and encourages use of public transit over single-occupancy vehicles. Well-designed public space leads to greater walkability in the area and safer environments that are more conducive toward active transportation for local residents. The daily movement also creates the potential for more socially interactive solar tracking if photovoltaic panels are installed on the canopy.

The shade canopy had to move from east to west to accommodate the daily movement of the sun, with the bench moving in the opposite direction to take full advantage of the moving shaded area. We explored a few options for using the bench itself as the actuator that rotates the canopy above. The rolling bench pushes a counterweight that offsets the weight of the canopy and allows users to easily adjust their position relative to the sun.

The movement of the frame was modeled and simulated in Grasshopper and Rhino to accurately place the counterweight and keep the frame balanced. The light-weight shade canopy was built with fishing line and fabric, and designed with parallel sheets woven between the lines that allow more sunlight through during winter months, and less during summer months.

This was a great opportunity to use our newly-available Lake Flato Shop across our parking lot and to utilize some of our in-house welding expertise (thank you Seema, Wayland, and Jamie!). We were able to use a few salvaged materials for our frame including an 8-foot long steel pipe as our counterweight and four yellow aviation pulleys as casters.

The frame can be disassembled into several parts for ease of transportation.  Once on site, two steel footings with rails are placed in position and two triangular steel frames are attached. A set of cross-bars with the fabric shade elements are attached to the triangular structure and tilted up into place. Our salvaged steel pipe is attached to the bottom of the rotating structure as a counterweight.  The bench is placed over the pipe, and acts as the actuator that moves the counterweight and canopy throughout the day based on the user’s preference.

The installation of the bench and shelter went relatively smoothly on Friday morning in our parking spot outside of 1221 Broadway. We recorded the movement of the canopy’s shadows in chalk as the sun and structure moved throughout the morning.

We learned a lot from our prototype and hope to refine its performance as it gets reused in different venues and reimagined as a permanent installation in Dignowity Park.  Our bus shelter produced a lot of good conversation and feedback, as well as the Most Creative Award from the Park(ing) Day judges.  The event was a great success and spurred a lot of dialogue and discussion about the possibilities of public transportation in downtown San Antonio. Thank you to Cotton, John, Heather, Serena, Alex, Seema, Catherine, Dan, Evan, Cecelia, Jamie and many others who helped us put together this fast-paced in-house installation.

Watercolor Wednesday

A few new drawings from Watercolor Wednesday, a weekly lunchtime watercolor class conducted by Matt Morris at Lake | Flato.

View fullsize

Lake | Flato's Josey Pavilion. The trees are black watercolor paint with a very dry brush, and the grass was done with charcoal pencil. Fun Fact: use a nail to make indentations for the grass, then go over the surface with a charcoal pencil. The vertical indentations will stay white.

View fullsize

A section through my Texas Academy of Science project. From right to left: entrance at Heritage Plaza, gift shop, courtyard amphitheater, planetarium, and gallery over the Trinity River.  This is white charcoal pencil on black foam core, with blue charcoal sky.

View fullsize

Lake | Flato's El Cosmico in Marfa, Texas, re-imagined as a monolithic stone ruin. A monument marking sunrise and sunset on the summer solstice.

View fullsize

A section perspective done with black watercolor and charcoal.

View fullsize

Fajada Butte at Chaco Canyon. Watercolor on black foam core.

One Year of Shirts

For the last year, I've recorded what color shirt I wear each day.

And the results are shocking.

No, not really.

Here they are, from May 1, 2015 to April 30, 2016.

6 white shirts at Chaco Canyon in May.

Less black in the summer, more black in the winter.

Much less color diversity after starting work in September.

114 days of dark blue (31%).  

179 days of any shade of blue (49%)

Bluest month: January (45% dark blue) (65% any shade).

Yes, I have more than one dark blue shirt.

Most consecutive days with the same color: 3 (on 6 occasions).

Next year: socks!

Energy Dots

Detailed home energy data is becoming more accessible with energy providers like CPS and PG&E replacing their electric meters with smart meters that log hourly data.  If you have hourly energy data and feel the urge to make a shiny rainbow calendar from it, here's how to make one with Grasshopper for Rhino.  This is seven months of my apartment energy use, with an energy dot every 15 minutes. The radius of each dot is relative to my total energy use at that time. The maximum value was about 1.8 kWh. The dot color represents the outside temperature, <40 degrees in blue, and >85 in red.

This is similar to an energy use graphic by Michael VanDaniker, but this simplifies the weather data from Weather Underground so you don't need to know how to use their API. 

The Grasshopper script, available for download here, takes a spreadsheet of energy data and a spreadsheet of weather data and plots out the energy dots in 3D.

If you have a smart meter installed by CPS, login to your dashboard, select My Energy Use > Access My Energy Portal, then Select view: by day.  Select 'Export your data' and 'Export usage for a range of days'. Open your .csv Excel file and delete all the title rows so your first line of data is in Row 1 like the image above.

Weather Underground has daily historical data available for download. Go to your city, select History > Custom, select the same date range as before, and scroll all the way to the bottom and select 'Comma Delimited File' to get hourly weather data as text. Copy and paste everything into Excel.

When your data is copied into Excel, make sure your data in column A is highlighted, select Data > Text to Columns > Delimited > Next > Check Tab and Comma > Next > Finish. Now your data should be separated into columns like the image on the right. Save your two spreadsheets as .csv or Comma Delimited files.

Now you're ready to import into Grasshopper!

The calendar view at the top is from the Top viewport in Rhino. Set the background to black by going to Tools > Options > Appearance > Color. Then export a hi-res image from Rhino by hitting Ctrl: P and selecting 'Image File' as the destination. Remember to check the 'Background Color' box under the Visibility tab.

After I export the image, I put it into Photoshop and add some Gaussian blur and Motion blur to smooth it out (Filter > Blur > Motion Blur, then set the right angle). Duplicate the original layer and set every duplicate layer to the 'Lighten' blending mode so they can be overlaid.

From the Front viewport in Rhino, we get a view of each day overlaid on top of the next.  Dots are vertically offset and colored based on outside temperature.  If you have even more detailed data from individual circuits, you can set the vertical offset to represent another variable like HVAC energy and get a nice energy profile of your building.

And the Perspective View is just for fun.