Search under noise
Multi-objective evolutionary search over constellation design spaces, trading delta-v, released dust mass, and remediation success rate under probabilistic constraints.
Virginia Tech · Aerospace Engineering
Truman DeWalch
Aerospace Engineering Ph.D. Candidate · Conjunction Analysis, Transfer Planning, and Mission Software
I build software for conjunction analysis, transfer planning, and mission design.
- Python, Rust, and C++ tools for orbital analysis and conjunction studies
- Led a 5-person systems team at The Aerospace Corporation
- Built the reusable event bank behind published astrodynamics work
Experience
Selected Experience
A mix of applied research, production-minded simulation software, and technical leadership in aerospace settings.
MTS Graduate Intern
The Aerospace Corporation · Northern Virginia / Remote
- Built a bistatic radar sensor model for an enterprise mission-analysis environment.
- Led a 5-person systems engineering sub-team across mission architecture research deliverables.
- Connected legacy tools so analysts could run high-level trade studies in one workflow.
Graduate Research Assistant — Just-in-Time Collision Avoidance
Virginia Tech · Blacksburg, VA
- Built Python, Rust, and C++ software for orbit determination and conjunction analysis.
- Developed a mixed-fidelity workflow that keeps early screening fast and uses higher-fidelity propagation where dust-cloud geometry matters.
- Compared optimizer families and deployer constellations under matched budgets, shared event evidence, and common scoring rules.
- Created a reproducible 1,000-event LEO conjunction bank for evaluation, benchmarking, and publication work.
Graduate Research Assistant — Eclipse Transient OD
Virginia Tech · Blacksburg, VA
- Designed a modular Python simulation architecture for eclipse-based autonomous navigation studies.
- Built a statistical atmospheric model from real climate data and used it in the estimation workflow.
- Applied Unscented Kalman Filters to estimate orbital state from eclipse-transient measurements.
- Presented the resulting approach at AIAA SciTech 2024.
Graduate Student Intern
TrustPoint GNSS · Northern Virginia
- Created high-fidelity models of satellites, ground stations, and onboard clocks for GNSS simulation.
- Implemented Extended Kalman Filters and Batch Least Squares for precise orbit tracking.
- Integrated TDOA, FDOA, and geometric-range measurement models into the simulation engine.
About
What I Build
Most of my work sits between astrodynamics research and the tools people actually use. At Virginia Tech, The Aerospace Corporation, and TrustPoint, I've built Python, Rust, and C++ systems for orbital analysis, led small technical teams, and published work in astrodynamics venues.
- 01
Build the models
Python, Rust, and C++ code for orbit determination, conjunction screening, and mission studies.
- 02
Make them usable
Turn research ideas into tools people can rerun, test, and trust.
- 03
Lead the work
Lead small technical teams and keep cross-functional projects moving.
Work
Selected Work
If you want the technical version, these pages show how the modeling and trade-study work actually fits together.
Simulation Framework
See Details
→ Simulated Space Environment
Propagation and analysis code for running orbital studies without rebuilding the stack every time.
Dissertation Research
Just-in-Time Collision Avoidance
Dissertation work on how a deployer architecture could respond to risky debris conjunctions.
See Details → Optimization FrameworkConstellation Optimization
Shared-evidence studies of how optimizer choice and constellation design change the trade space.
See Details → Approach
Methodology
These are the pieces I keep fixed so the comparison stays honest across optimizers and constellation studies.
- 01
Reusable event evidence
Catalog-anchored hazardous-event set reused across optimizer and constellation studies
- 02
Two-lane dynamics policy
Screening stays analytical while released dust gets high-fidelity propagation
- 03
Objective / gate separation
Released dust is optimized while intercepted mass remains a strict feasibility gate
- 04
Fair optimizer comparison
Matched budgets, shared event draws, and common scoring keep comparisons interpretable
Optimization
Optimization Background
My dissertation is a stochastic optimization problem end to end: population-based search over constellation designs, where every objective is a noisy Monte Carlo estimate and every candidate competes under a fixed evaluation budget. The machinery generalizes well beyond astrodynamics.
Sample-efficient evaluation
Objectives are Monte Carlo estimates on a budget, so runs use adaptive evaluation policies and early stopping to spend samples where they change decisions.
Benchmarks before conclusions
A reproducible conjunction-event bank, built by surrogate sampling of the debris catalog, keeps optimizer comparisons on shared evidence and common scoring.
Fast objective functions
Rust batch evaluators and mixed-fidelity propagation keep single evaluations cheap enough that large searches stay tractable on realistic physics.
Where it applies
Machine learning & AI
The problem structure behind hyperparameter optimization and Hyperband-style schedulers: black-box search over noisy objectives with budgeted, racing-style early-stopped evaluation — plus the benchmark and evaluation-harness discipline ML teams rely on.
Probabilistic modeling
Sigma-point propagation, Gaussian mixtures, and divergence-based validation are shared vocabulary with probabilistic ML and uncertainty quantification.
Engineering design
Architecture trade studies — from sensor placement to constellation geometry — are design-space exploration under constraints: the same optimization loop in a different domain.
Publications
Publications
Conference papers and presentations from the dissertation thread, including the papers where I was second author. If a direct link is missing, the citation search is the quickest way in.
Lead-author work
The papers where I led the framing, writing, and conference presentation.
2026
Stochastic Optimization Techniques for the Design of Just-In-Time Collision Avoidance Constellations
Read Paper
Direct paper
2024
Enhancing Eclipse Transient Orbit Determination Methods with Statistical Atmospheric Models
Read Paper
Direct paper
2024
Dispersion of Targeted Orbital Dust Clouds: Applications to Just-in-time Collision Avoidance
Find Citation
Citation search
Collaborative work
Closely related conference work where I contributed as second author.
2025
Statistical Evaluation of Dust-Based JCA Systems and Policies
Find Citation
Citation search
2023
Orbit Determination via Eclipse Transient Timing: Improved Methods and Intensity Models
Find Citation
Citation search
Skills
Technical Snapshot
A quick scan of the languages, methods, and tools I use most often.
Programming
Python NumPy, SciPy, scikit-learn C++ C Rust MATLAB Fortran Julia
Technical Areas
Kalman Filters UKF, EKF Orbit Determination Conjunction Analysis Evolutionary Algorithms Monte Carlo Simulation Statistical Atmospheres Uncertainty Propagation
Core Tools
Git Linux Unix STK Level 3 Certified Jira Confluence LaTeX
Education
Education
Ph.D., Aerospace Engineering
Aug 2022 – Sept 2026 (Expected)Virginia Tech
Focus: High-fidelity OD, Conjunction Risk Modeling, Statistical Atmospheres. Advisor: Dr. Riley Fitzgerald.
B.S., Aerospace Engineering
Aug 2018 – May 2022Virginia Tech
Certifications
STK Level 3 Certified
Leadership
Leadership & Activities
Rank Captain & Tuba Player
Aug 2018 – Jan 2025Marching Virginians
- Led a section of 24 members, teaching music and marching fundamentals while managing logistics.
- Organized and participated in service projects, including the "Hokies for the Hungry" food drive.