Sunday, November 15, 2015

Martian Mesa

I have a fairly lengthy commute to and from work. It was one of the only drawbacks to when I accepted the job. However since I have started that commute I have discovered some most excellent podcasts which now fill every minute of that commute and sometime even more. Most of my stations focus around coding, space, statistics (within the realm of data science), robotics and almost anything science related. With this bit of background my projects list might makes a little more sense.

Mars or Mesa

While cruising into work one morning a TED talk came up on my playlist with a guy named David Lang. David taught himself how to build a robot and start collecting oceanographic data. How he got started out was a lot like how I am trying to get going, he needed to collect data, didn’t have the money to buy the robot or equipment needed to do it, so he built it. He later went on to create openrov where he sells the kits he came up with for super cheap (comparatively). He got talking about collecting data remotely and using robots to do it and it totally inspired me to get back to my robot projects.  That was on the way in to work so I spend a good deal of time pondering what the next step was for my weather station and data collection units and how to make it mobile. Then on the way home I listened to one of my Space focused podcasts. 

With the recent release of the movie the Martian, which I loved, a lot of my space focused stations spent a few episodes discussing Mars exploration. One in particular had a panel of folks who represented specific areas within science who were interested in space exploration. There was a great discussion on whether we should focus on sending astronauts and scientists to Mars or if we should just keep sending robots. There was a geologist who said something a robot would take weeks to analyze, or possibly could never do, he could do if he were physically in the space in a couple minutes. The roboticist of course was like no way, I could get it done and then the astronaut was like hell no send me up and then the guy from NASA who obviously represented the budget was like well this is all going to cost  a lot of money. It was a great discussion and it inspired me to want to get into the space race and robotics. Then these two podcasts came together in my mind and my imagination almost exploded.

After these two came together I instantly wanted to build a robot that could roam around and start collecting data. This thought occurred to me right as I was cruising the North Shore Road coming around Lake Mead and I saw the Mormon Mesa and figured that it would be a perfect testing grounds. The Mesa Rover was born.

As I started to research the Mars rovers and look at pictures they had taken I was impressed with how similar Mars and the Mormon Mesa looked. The boys and I convened a conference in the shed and it was decided to begin work immediately.

So far we have built many protypes but feel like we have found a good design for the tracked robot. We have a pretty good center of gravity so that if it goes back onto the real part of the tracks or above 38 degrees pitch then if you put it in reverse it will go back onto the main part of the tracks.

We have also tested out the battery life and it will go .8 miles in 56 minutes on 8 recharable 1.2 volt 1300 MAH AA batteries. There is still a lot of testing to do with adding on the solar panels and recharging the batteries but it’s good to have a base line to run off of.

We also have the sensor pretty well dialed in that gives pitch, roll and heading. Today we calibrated the sensor to maintain a heading between 264 and 276 degrees which is a nearly easterly direction which is heading for the road that runs by our house. It took us a little bit to find the right course correction algorithm so as to not spend too much time bouncing back and forth between those headings.

Our next step will be to add in the sonar for object \ cliff detection. I will keep update as we make progress on our Mesa Martian Rover.

Hubble Telescope and Gold Butte
The next project has been in the works for a long time but I finally was able to bring it together after listening to one of my python coding podcasts a couple weeks ago.

Ever since I did the fire analysis project in 2014 I have wanted to study the effects of wild fires in the desert. I wanted to better understand how the desert responds after a fire but also possibly identify areas that haven’t yet burned but are most prone to so I could focus research there so if a fire did happen I would have a good base line to run on pre-fire. One of the pieces of analysis I would like to be able to do is to quantify vegetation cover or counts of plants per sq meter in a particular area. To do this I would need to be able to perform somewhat complex remote sensing type analysis. There is software out there that can do such analysis but I don’t have it nor the experience to run it, so I figured writing my own software would be a more attainable approach for me.

While driving into work a couple weeks ago I listened to a podcast called “Talk Python To Me” the host interviewed a scientist who was using part of the Science Kit modules called SKImage. The scientist was using the software to write algorithms to count cells in an image from a microscope but in a high level overview it’s the same concept as I wanted to do with aerial photographs of the desert. After listening to this it sounded like this would be a possibility for me to start work on my image analysis tool.

While researching the skimage module I found an algorithm that a team had written to analyze images from the Hubble Telescope. While looking at the images from the Hubble I thought that they didn’t look too different from a desert landscape found in Gold Butte or Mormon Mesa. I was able to take that program that they wrote and modify it to analyze aerial photographs of our beautiful desert. I am still trying to hammer out the best variables for the formulas but I have a pretty strong start.

So how does this all relate to gold Butte you may be wondering? Well the mesa is not that far from my house so it makes it an easy test ground to run my experiment and it is also adjacent to Gold Butte so once I get the rover running on the Mesa I can then send it out into the wilds of Gold Butte to start collecting data. What kind of data am I going to collect? All kinds of good stuff. Really anything you can think of there are sensors to detect and measure.  I will start with weather data and then add in elevation, slope, available light, wind and all kinds of other factors so I can start building a model to more accurately forecast and predict weather patterns on a microclimate level. I would like one day to be able to integrate the camera as well as an infrared camera to start being able to detect vegetation and write a detection and classification program to start classifying and quantifying vegetation and start building GIS data. I also have ideas for a habitat mapper\monitor\data gathers but I am already getting way ahead of myself.  I also would like to have the bot be able to explore and map abandoned mines. I have directional sensors, sonar, gas detection and other air quality sensors for this project.  In the mean time I am still dreaming things up so I will post my ideas as I make progress.

Wednesday, November 4, 2015

Lessons Learned and Up Next - DRASTIC

Well I finished up my last weather project. You can see the results here . For that project I created two weather stations and compared the temperature on the Virgin Mountain at 7,100 ft. and in Logandale at my house at 1,400 ft on the same line of latitude. The results of the project were that I found the temperature difference was 24.6 degrees or roughly .004297 per foot of elevation gain. It seems kind of sad to summarize the whole project up in one succinct sentence or really one number. This project was fairly involved and actually took me a couple different tries running up and down the mountain over the course of a month to collect the data. Storms, hardware failure and technical difficulties were plentiful. However there were multiple returns from the project, including tangible data as well as knowledge gained with hardware and software. These were huge wins for me that will benefit future projects by streamline the time it takes to implement those projects. The knowledge I get to carry forward really is the biggest result of this last project.

The knowledge gained from the project was not just that the temperature is 24.6 degrees cooler on the mountain than at my house but in building the systems to collect the data, the software built to analyze the data as well as a better understanding of the weather. I also had hardware outputs from the project that I can use in future projects. The building of the systems was also an endeavor in and of itself. I needed to be able to collect many data points for an extended period of time without the ability to plug it in while also being modular enough that I could pack it up in my backpack and hike it to remote locations and have it withstand the elements. The actual building of the hardware components was not a huge task because I had previous experience with micro-controllers and soldering so I was able to spin those up pretty quick. My weekly allowance for sensors held me back more than anything. The part that took me probably the most to hammer out was the power supply.

There were two parts to this, one: writing code that minimized the amount of power used by the micro-controller and sensors and two: Creating enough power to supply the station with adequate power as well as charge the batteries that would allow the station to run during the night when solar power wasn't available. This wouldn't really be a huge deal except for the size of the panels and for money. As I mentioned earlier I needed the station to be as small as possible so I could pack it in my backpack as well as being inconspicuous so people would leave it alone. I am also funding all this on my own dime so I am going as cheap as possible because it cuts into my soda fund. After much testing I was able to reduce the power consumption considerably and create a power supply unit that was small and reasonably priced for a backyard hacker.

This left me with the last piece which was to weather proof the unit so it could withstand fairly intense weather. This took some effort and testing as well as a few failed attempted deployments on the mountain. However finally I was able to get the two stations to run in parallel long enough to collect data and mark the project a success.

(my bed out in Cedar Basin)

For my next project I will be adding a few new components into the monitoring system but I am also shifting the focus of the data I am going to gather. One of the best parts about getting out and about is seeing wildlife. I love when you’re cruising up the mountain and you jump a few deer up and get to watch them bound up the steep mountain slope. I love when you’re out in the hills and you catch a glimpse of a coyote sleeking across the hills and valleys. I enjoy watching the Jack Rabbit bounce between the cactus and sage or the Gamble Quail scurry across the hill side in the waning afternoon sun. The focus of my next project will be to gather data about wildlife and their habits, potential habitat areas and the interactions between the wildlife in the shared ecosystem.

One of the new pieces of hardware that I ordered last night was a camera that interfaces with the Arduino micro-controller system that I use. I also have a motion detecting sensor so I will wire these two pieces onto the system so I can take a picture when there is movement. I also have built a wind monitor so I can try to better detect if the motion is the bushed blowing in the wind or an actual animal moving within the field of view of the camera.

With these additions to the system I hope to be able to capture some good pictures of different wildlife out in the Gold Butte region. I am essentially building a trail cam but I will also have all my other sensors hooked into the system so I will be able to measure temperature, humidity, barometric pressure, wind speed, light, date and time and soil moisture. I hope with all of this data that I can start to get a better understanding of particular animal behaviors and start to find trends and habits.

The animals that I am going to focus on are Gamble Quail, Jack Rabbits, Coyotes, Mule Deer and Mountain Lion. While I am waiting for me new hardware pieces to arrive I have begun the process to define the area where I will set up my monitoring station. I started first by downloading all of the Spring GIS data for the Gold Butte region. I then went to the Nevada Department of Wildlife (NDOW) website and downloaded their Mule Deer habitat GIS data. I then overlaid the landform dataset which gives classifications of slope as well as a vegetation dataset. With data I began the process of finding the most suitable areas where wildlife would potentially be found.

After analyzing the data I have identified the area to where I think I will have the best probability to capture data on the animals I outlined previously. After I get all of my hardware assembled and tested I will deploy my new station near Gold Butte headquarters.

There are many different types of analysis that I will be able to perform with the data that I hope to collect. One dataset that I will hope to be able to extrapolate will be species population counts. If I monitor different springs and get good counts of Quail that come and go throughout the morning and evening hours and these counts are roughly the same I will be able to estimate how many birds reside within a given area. I will also have to take into account the terrain and vegetation but I have this data and will be able to create a fairly sophisticated model to estimate this data. If I can get counts of deer and know their normal patters for watering and know their preference for feeding based on available forage I can start to understand where they will most likely to be found and when and under what weather conditions. If I can predict where quail and Jack Rabbits will be then I can start to paint a better picture of where and how many coyotes reside within a certain region. I also hope to be able to capture pictures of a mountain lion out in the Gold Butte region. I have never seen one personally but know people who have and I would like to see if I can get a picture of one and hopefully multiple pictures at different springs so I can start to understand mountain lion habits as well.

This is going to be an ongoing and long term project. Just getting the new hardware pieces integrated into my existing system and tested and weather proofed is going to take some time. My current knowledge and understanding of these animals is quite simple at this time. To help me kind of get a leg up on it all and up to speed I have enlisted the help of a master woodsman and animal tracker, and longtime friend, Kyle Leavitt. He has already donated some time in getting me up to speed on how to track and monitor mountain lions and mule deer. He also has a firm grasp of how to utilize scents to attract these animals and will help me to increase the odds of getting my first data points with my monitoring system. I will post updates as I move along on this project.

This is a time lapse of the Logandale Trailhead with a picture taken every ten minutes between 9 AM - 5PM. The Camera used was an AdaFruit TTL camera that was run from my Arduino which was powered by solar. The video isn't that great but it was a big accomplishment getting the system to run so here it is.