For the last 3+ years, I have done Guidance, Navigation, and Control work at Raytheon in their Missile Systems branch. Here, I primarily helped build algorithms to make weapons smarter, make predictions, and design other sophisticated numerical algorithms for data analysis and performing Design of Experiments. For some other related projects, I also did unique work like building mobile mission planning/simulation applications for soldiers and tackling inverse problems for the US Navy or US Air Force.
Starting Fall 2017, I will be a full-time graduate student in Computer Science @ University of Illinois at Urbana-Champaign. My exact focus has not been fully decided but my interests currently lie in Artificial Intelligence, Scientific Computing, and CS Theory.
Outside of work, I am passionate about math, physics, computer science, and application to areas of simulation, controls, optimization, and AI. I enjoy reading about and working on projects related to the above topics, with more recent efforts being in mathematics, computer science, and reinforcement learning. I am also a cinephile and have a great interest in story telling via written and visual mediums. Oh and I have a dog that I love who keeps me grounded in reality!
The simplest route to reach me is via my gmail account where my handle is choward1491.
A more recent project at Raytheon I have made great progress on was building a framework for distributed optimization, performed using heuristic optimization schemes like Particle Swarm Optimization, that can be used to tackle optimization problems with computation heavy and time consuming cost functions. With this framework, I integrated our Six Degree of Freedom (6DOF) simulation, that models weapon engagement scenarios, to create an optimally tuned set of Bayesian estimators for estimating target motion. What once took months for a person to hand tune has been turned into a job that takes up to a couple days to achieve performance superior to that of the human tuned baseline.
I have also used this framework to begin progress in crafting an optimal guidance algorithm, based on a parametric formulation using Neural Networks, for producing a weapon that can hit a target at some designated time. This latter task was never completed while I was at Raytheon Missile Systems, but my work provides a foundation upon which other GNC engineers can work upon to building a superior guidance algorithm.
Another project I built in my time at Raytheon was a Domain Specific Language, named tesl, used to compile scattered datasets on our cluster together and fuse them into a single dataset that could be integrated into other codebases we used (like a library for producing high fidelity Launch Acceptability Regions). This codebase modeled the scattered datasets as a hyperrectangle tessellation and used this assumption to find hyperrectangles that enclosed a desired data point using a fast data structure I designed. Other selectors were also written in case various subsets had different results for the same high dimensional coordinates, allowing us to fuse the results using various filtering techniques. Under the hood, the overall set of coordinates would be represented as a dense graph and would be traversed based on a pattern set in the input script.
Using this interpreted language, written in C++, we were able to efficiently compile many datasets together and greatly reduce time needed to get from generating data to using it in our libraries.