Project Grizzly

Inspiration

We were inspired and motivated by the countless forest fires in California. The number of forest fires has been increasing consistently every year in California. Most solutions so far are off ground solutions. These projects don't gather data from the ground where fires actually take place. Our goal is to use environmental data and animal behavior to predict and locate potential areas where forest fires can start. The name Project Grizzly comes from California's state animal, which is a grizzly bear. California is an example location of where our network could work.

What it does

Project Grizzly is not a single IoT device, it is a network of data gathered by smart devices and then processed by machine learning algorithms. Here is how everything is connected:

Project Grizzley uses wildlife behavior to track patterns in movement to locate potential fires starting. Large mammals such as elk, deer, and grizzly bears are known to not only act quickly but also act before fires become largely detectable. Inserting GPS tracking systems into animals allows us to view when a change in their behavior occurs. For example, elk and deer run towards bodies of water when fires break out. If Project Grizzley’s algorithm sees a large population of deer and elk moving towards bodies of water, then there is a strong change a certain portion of the forest is engulfed in fire.

Project Grizzly uses five foot tall rod to detect biomass density and environmental changes to assess potential outbreaks of fire. Forest fires feed off of dense leaf and plantae coverage. Knowing which portions of the forest have a growing collection of dead leaves and thick moss allows us to know where to be wary of forest fires. The ROD gathers this data using… It then sends this data via satellite using satellite networking solutions such as Abela by Inmarsat. This data gets processed by Project Grizzly and then is determined to be potentially dangerous or not.

Data acquisition and analysis is what makes Project Grizzley so powerful. The use of various data sets along with machine learning algorithms, allows Project Grizzley to predict and locate potential forest fires in continuously changing ecosystem. Project Grizzley uses MongoDB to store data from the ROD and GPS tracking systems. This data is then processed through IBM Watson to get probability values of forest fires occurring in specific locations.

Project Grizzly adjusts and gathers data for unique environments. Every ROD and tracking group gathers data for a specific location, which allows us to monitor precise areas.

How we built it

Essentially our data from our GPS tracking system and ROD device is sent to our MongoDB server using the satellite technology, Abela by Inmersat. This data is then processed by our model in IBM - Watson.

Challenges we ran into

The biggest technical challenge we ran into was using IBM - Watson to run data processing tasks on our data set. The next challenge was product design. We spent a lot of time trying to figure out was Project Grizzly is and what exactly it does.

Accomplishments that we're proud of

Going off of the challenges we faced, we are very proud of the vision of Project Grizzly and what it is designed to do. We think our network design is unique and well thought out. Our intended network is complex but solves its intended problem with sensible solutions. Tracking animals and having a biomass and local environment sensor makes a lot of sense.

What we learned

We learned a decent amount of MongoDB and IBM Watson technology. We realized a new, at least for us, usage for both of these technologies. It was also the first time a couple of us tied together data acquisition with machine learning.

What's next for Project Grizzly

The next step immediate step for Project Grizzly would be to use a prototype ROD to gather biomass and weather data from a local environment. We want to then process this data and see what we can model from data processing. We would also like to gather GPS data for animal patterns and run it into machine learning algorithms like IBM Watson. Though, this animal tracking step is a little more resource heavy than a prototype ROD. If we were to carry out this project down the road, we believe that a sufficient prototype network consists of tracking systems in a small population of grizzly bears, elk, and deer along with dozens of ROD devices placed in different parts of forests. At this point even tho we would not have much data, we would want to still process it in relatively real time through a satellite and into our machine learning model.