Group Simulation within Position Based Dynamics Crowds
Summer 2022
I joined Professor Tomer Weiss’ lab at NJIT, were I worked to better simulate groups within Position Based Dynamics (PBD) crowds. I created a visualization in the Unity Game Engine to display results, created a novel method of agent assignment to improve group reformation, and wrote significant portions of a paper.
Urban Heat Island Investigation at NASA SEES
Summer 2022
I participated in the 2022 NASA SEES high school internship program, where I worked with a mentor to learn about climate science, remote sensing, and the Urban Heat Island Effect (UHIE). I worked with a partner to collect satellite data and analyze it, examining if there was a correlation between the amount of groundwater and an area and the UHIE. I used Python to download Landsat-8 imagery of 100 large American cities for analysis. Using the Artificial Surface Index (ASI), I identified urban and suburban pixels in each image and calculated the average temperature of each from the Landsat surface temperature band. By examining the difference in average temperature, we measured the UHIE at each city. Groundwater data was retrieved from the GRACE dataset. Ultimately, no correlation was found, though I learned a lot from the experience. I presented our work at the SEES Science Symposium (video here).
Stadium Crowd Simulation
Winter 2021/2022
This project was inspired by the 2022 TSA TEAMS competition. Despite the fact that the original assignment was to propose a solution to optimize the stadium concert experience, I decided to take the opportunity to do something more like real research. I found a floor plan of a typical stadium and created a simple image representation scheme for the seating areas, walls, and exits. I then wrote a python script that simulated individual agent behavior based on this paper. I tested 4 different potential solutions against a control condition, analyzed the efficacy of each using the simulation, and made a recommendation to improve crowd flow.
The source code is available on GitHub.
The final paper is available as a pdf.
Landsat 8 Pine Island Glacier Visualization
Fall 2021
In this project, I used the Machine-to-Machine API with Python to search the Landsat 8 OLI/TIRS tier 2 database with the scene-search API endpoint. Metadata filters were used to select only scenes of a specific location; the Pine Island Glacier in Antarctica. The imagery was of an 185×180 km area. The download-options endpoint was used to download these scenes with Python. To avoid heavy cloud cover, all scenes with a CLOUD_COVER greater than 25 as read from their MTL metadata file were ignored. This had the side effect of only selecting images from the summer months (October – February). The remaining scenes were then processed using the GDAL Python library to read the GEOTIFF format and extract bands 2-4. These bands were then combined into a RGB image using numpy, and outputted using matplotlib. The images were collected with OpenCV to generate a final mp4 file showing the state of the Antarctic Pine Island Glacier from 2013 to 2021.
The Antarctic Pine Island Glacier is one of the fastest melting glaciers in the world, something which is actually visible in the imagery – simply comparing the first to the last frame, it is evident that a substantial chunk of the glacier body itself (center top) has disappeared.
The source code is available on GitHub.
Ozone Data Visualization
Fall 2021
This was essentially a feasibility study to see if I could extract satellite information from the USGS database. I used the bulk download feature of EarthExplorer to collect ozone data from the MODIS Fused C2 dataset, then used matplotlib as a visualization tool to graph the data in Python. I then collected the data with OpenCV to generate a final mp4 file showing the ozone layer during every month from 2013 to 2021.
The source code is available on GitHub.
