top of page

Current Research and Opportunities

My NEural TransmissionS (NETS) research group consists of 3 Ph.D. students, 2 M.S. students, and 6 undergrad students, although we collaborate with many others. 

I have some projects with opportunities for students to study, so I have summarized the state of my open projects below (as of May 2022)

Machine Learning

This research area has consumed most of my interest in the past couple of years, through teaching, research, nonprofit, and industrial work.

  • Aerospace Engineering: Deep learning for computer vision and in-orbit satellite component detection to support in-orbit servicing and space debris-capture missions.

    We have implemented state-of-the-art object detection algorithms to automatically classify and localize objects like solar panels, antennas, cubesats, and satellite bodies and run tests in Florida Tech's ORION Research Lab.

    Currently, we are tuning the models, developing best practices to deploy models on heavily limited computational resources characteristic of what can be implemented on small chaser satellites.
    [Slides from a recent research presentation]
    [Project supported by the U.S. Space Force and Air Force Research Lab.]

    T. Mahendrakar, A. Ekblad, N. Fischer, R. T. White, M. Wilde, B. Kish, and I. Silver (2022). Performance study of YOLOv5 and Faster R-CNN for autonomous navigation around non-cooperative targets. 
    IEEE AeroConf 2022.
    T. Mahendrakar, J. Cutler, N. Fischer, A. Rivkin, A. Ekblad, K. Watkins,
    R. T. White, M. Wilde, B. Kish, and I. Silver (2021). Use of artificial intelligence for feature recognition and flightpath planning around non-cooperative resident space objectsAIAA ASCEND 2021.
    T. Mahendrakar, R. T. White, and M. Wilde (2021). Real-time satellite component recognition YOLO V5. 35th Annual Small Satellite Conference.


  • Novel Object Tracking Approach: To support the satellite project and beyond, we are developing novel object tracking algorithm that can run on edge hardware.

  • 3D Pose Estimation: Another group is working on computer vision-based estimation of the 6-DOF pose (position and orientation) of an object in three dimension based on a camera feed. See some initial work on spacecraft pose estimation.

  • Glaciology: Another group is developing an approach to track the evolution of glaciers over time using time-series satellite imagery. After testing on some manually-gathered data, we developed a data pipeline to construct large datasets of images along with various variables associated with each glacier at each time. We are currently tuning neural image segmentation algorithms to automatically segment the glaciers in multispectral satellite imagery. See some recent developments by the team.
    [Related work
     supported by the National Science Foundation in the SMAG REU Program]
    [Current satellite data access provided by the European Space Agency]


  • Global Development: I serve as Senior Advisor on Data Sciences to non-profit organization Engage-AI on a research program to leverage machine learning and artificial intelligence to find best practices in pursuing development goals, such as the UNDP's Sustainable Development Goals (SDGs). The work involves developing an AI-enhanced data platform drawing from disparate data sources to facilitate data analysis, finding useful patterns in the relevant data with state-of-the-art machine learning methods, and working with governments and NGOs to shed light on pressing development issues.

  • Intrusion Detection: I am interested in pushing the boundary of the sorts of intrusions we can detect by taking known attack signatures attempting to learn novel vulnerabilities by generating variations of known signatures via convolutional autoencoders and generative adversarial networks.

  • I am also open to supervising reasonable machine learning projects proposed by students under certain circumstances, especially those involving neural methods. Here are some examples.

Stochastic Processes

Most of my published academic work is in the area of stochastic analysis and probability theory. I have many projects ongoing in this area, which I have broken into several categories, although the categories have some overlapping content.

  • Reliability of Stochastic Networks: I'm studying another random process taking place on large weighted graphs (networks) where, at random times, random batches of nodes get incapacitated, each with a random number of edges with random weights. I'm interested in finding how long it takes for the sum of nodes or edges or weights lost to surpass some given thresholds.

    I have mathematical results, but the next step is to simplify the results to more practical situations, which involves finding the inverses of certain operators, probably via numerical approximations, to confirm they are easy enough to compute to be useful. Then, simulations are needed to confirm the formulas match empirical results.

    R. T. White (2015). Random Walks on Random Lattices and Their Applications. PhD thesis, Florida Institute of Technology. [slides] [fulltext]
    J. H. Dshalalow and R. T. White (2014). On Strategic Defense in Stochastic Networks. Stochastic Analysis and Applications, 32:3, 365-396. [arXiv preprint]
    R. T. White (2013). Stochastic Analysis of Strategic Networks. 38th Annual SIAM Southeastern Atlantic Section Conference. Melbourne, FL. [slides]
    J. H. Dshalalow and R. T. White (2013). On Reliability of Stochastic Networks. Neural, Parallel, and Scientific Computations, 21, 141-160 [arXiv preprint]


  • Random walks: these are points moving around in n-dimensional space by making random jumps at random times. I study the dynamics of these processes when they exit from a fixed set in the space.

  • Minimal Sufficient-Probability Sets: I am studying the evolution of small high-probability sets for the location of a stochastic process. We study how these sets change over time by watching how probability density flows through the boundary of the sets from the previous moment in time. I aim to continuously deform the sets over time by analyzing the flux across their boundaries.​​​ Much mathematical and applied work is needed.

    • Mathematical: expand beyond toy examples to determine more general conditions under which it can be done and to find closed-form solutions in these settings, if possible. (It should heavily relate to PDEs, but I have not investigated the link well.)

    • Applied: implement mathematical results in practical problems, write efficient algorithms to compute similar results for simulated processes where closed-form solutions are elusive, and ensure acceptable error bounds.

R. T. White (2020). On the Evolution of Minimal-Volume, Sufficient-Probability Sets for Stochastic Paths. 14th International Conference in Monte Carlo & Quasi-Monte Carlo Methods in Scientific Computing, Oxford University. [slides]


Statistical Models with Applications to Geoscience REU Program

During the summers of 2021-2023, FIT hosted an NSF-supported research experience for undergraduates (REU) program involving projects in climate science and marine biology. It was a  great opportunity for aspiring scientists to participate in a funded program, get some training, and gain research experience.

Students: please feel free to get in touch!

bottom of page