Aerospace Engineer · Riyadh

Hamoud Alshammari

I work across astrodynamics, spacecraft estimation and Earth observation, from orbit determination and attitude control to SAR, LiDAR and gravity fields, building tools that are tested well enough to trust.

M.S.E. AerospaceUT Austin
Space researchCenter for Space Research
IndustrySatcom · PNT · Earth observation
Earth seen from space over Europe, Africa and the Middle East, with satellites in low Earth orbit
LEO shells · 550 km 53° · 700 km SSO · 1200 km 88°

01 · About

I turn measurements from space into knowledge we can trust: where a spacecraft is, where it's pointing, and what it sees.

I am an aerospace engineer based in Riyadh. I hold an M.S.E. in Aerospace Engineering from The University of Texas at Austin and worked at UT's Center for Space Research on lunar geodesy and reference frames. Since 2025 I have worked in industry on satellite communications, navigation and Earth-observation systems.

My work spans three connected areas of space engineering. What ties them together is the same habit: model the system, estimate it from real or simulated measurements, then test the result hard enough to know how far it can be trusted, whether that means validating against a reference implementation, running a Monte Carlo, or checking a covariance against the actual error.

01 · Astrodynamics

Orbit determination, design & control

Estimating and predicting orbits, designing them for a mission, and keeping them there: perturbation theory, propagation, maneuver planning and maneuver detection.

EKF/UKF orbit determination · TerraSAR-X orbit control · lunar halo orbits & Mars transfers · SGP4 & propagators · celestial mechanics

02 · Estimation & autonomy

Navigation, attitude & spacecraft estimation

Knowing a spacecraft's position and attitude on board: Kalman and particle filters, multiple-model methods, GNSS navigation and integrity, and attitude determination and control.

non-Gaussian orbit uncertainty · navigation on a lunar halo orbit · ADCS with MEKF · 37-filter Julia toolkit · GNSS integrity (RAIM) · optimal control

03 · Earth observation

Remote sensing & satellite geodesy

Turning satellite data into information about the Earth and Moon: SAR imaging, LiDAR, machine learning on imagery, gravity fields and reference frames.

SAR back-projection · tree-crown detection · LiDAR · gravity-field modeling · lunar frames at CSR

02 · Experience

Experience & education

May 2025 – present

Aerospace Technology Engineer

Neo Space Group · Riyadh

R&D and technical due diligence across satellite communications, positioning/navigation/timing, Earth observation and NB-IoT. Much of the work is checking whether a quoted performance number still holds once its assumptions are written down, through link budgets, coverage models and constellation-geometry (dilution-of-precision) studies.

Feb 2024 – Apr 2025

Research Engineer, Graduate Research Assistant

Center for Space Research, The University of Texas at Austin

Lunar geodesy and reference-frame research in a collaboration between CSR and CU Boulder's Colorado Center for Astrodynamics Research. Reviewed the literature on lunar frame realizations and frame-tie strategies, evaluated how existing multi-source data products could support a consistent lunar frame, and co-authored internal technical documentation of the team's assumptions.

Research Engineering/Scientist Professional (Jan–Apr 2025) · Graduate Research Assistant (Feb–Dec 2024). Project details are limited because the work was internal.

2023 – 2024

M.S.E. in Aerospace Engineering

The University of Texas at Austin · GPA 3.9/4.0

Orbit determination, satellite geodesy, estimation, attitude control, celestial mechanics and remote sensing. Coursework →

2018 – 2020

B.S. in Engineering Mechanics

University of Wisconsin–Madison · Aerospace focus

Dynamics, vibrations, experimental and computational mechanics; senior design on a reusable lunar lander. Projects →

03 · Projects

Selected projects

Recent independent work, then graduate projects from UT Austin. Click any figure to enlarge it.

Recent work independent · 2026

Estimation · uncertainty propagation

When one Gaussian is not enough: orbit uncertainty after a tracking gap

Added an adaptive Gaussian mixture filter to my Julia estimation toolkit. It starts from a single Gaussian and splits a component in three when the dynamics become too nonlinear across it (an entropy criterion after DeMars, Bishop & Jah), keeping the mean and covariance of the mixture exactly. I tested it on a 700 km satellite left untracked for two revolutions (3-D, two-body + J2), scoring each filter against 20,000 Monte Carlo samples, then through a radar pass in 200 trials. Writing the tests also exposed a Gaussian sum filter that had never worked and a likelihood underflow in nine particle and mixture filters; both are fixed.

Gaussian mixturesEKF / UKFKL divergenceMonte Carlo · NEES

Outcome: after the gap the KL divergence from the true distribution is 2,700 nats for the EKF, 19 for the UKF and 2.5 for the mixture. Through the pass the mixture stays consistent (92–97% of trials inside the 99% bound, against 8% for the EKF and 34–63% for the UKF) and is 3 to 9 times more accurate than the UKF. Code →

After two revolutions without tracking the uncertainty is a curved band several hundred km long; the EKF and UKF ellipses do not follow it, the 49 mixture components do. KL divergence from the true distribution during the gap: thousands of nats for the EKF, tens for the UKF, a few for the mixture. Radar pass after the gap: the mixture's NEES sits at the chi-square median while the EKF's is near 100,000; the mixture's position error falls below 300 m.

Astrodynamics · cislunar & interplanetary

Lunar halo orbits and Mars launch windows from first principles

Rebuilt my Julia orbital mechanics package around the circular restricted three-body problem: Lagrange points, a differential corrector for periodic orbits and pseudo-arclength continuation. From the Earth–Moon mass ratio alone it finds where the L2 halo family branches off the planar Lyapunov orbits and follows it down to the near-rectilinear halo orbits (NRHOs), including the 9:2 resonant NRHO chosen for the lunar Gateway. The same package maps Earth–Mars launch energy over departure and arrival dates. Testing against textbook examples also exposed errors in my original Lambert solver and J2 propagation, now fixed.

CR3BPHalo orbits / NRHOContinuationLambert

Outcome: the 9:2 NRHO comes out with a 6.562-day period and a 3,249 km perilune, its initial state within 1.5 × 10⁻⁵ of the published one; the corrector reproduces Howell's 1984 halo orbit to six digits; and the 2020 Earth–Mars minimum falls in late July 2020, when Hope, Tianwen-1 and Perseverance launched. 516 tests. Code →

The Earth–Moon L2 southern halo family from its bifurcation down to the NRHOs, with the 9:2 NRHO highlighted; stability index and period along the family. Earth–Mars launch energy for the 2020 and 2026 opportunities; the three 2020 missions sit in the low-energy valley.

Navigation · cislunar

Navigating the Gateway's halo orbit: where the EKF is right

Joined the two packages: an EKF, a UKF and the adaptive mixture tracking a spacecraft on the 9:2 NRHO from one ground station, with the orbit and its state transition matrix from the astrodynamics package. A three-revolution coast scored against Monte Carlo, then 100 tracking trials with a 3.5-day station outage across a perilune.

NRHOEKF / UKF / mixtureCovariance realism

Outcome: the opposite of the low-orbit result. Away from perilune the EKF's density is exact even after three revolutions, while the UKF and the mixture drift from the truth and do not recover; at perilune the EKF is badly wrong for a few hours, then right again. The mixture's split criterion cannot tell a bend that undoes itself from one that does not, a limitation the write-up spells out. Write-up →

Three revolutions of the NRHO without tracking: the EKF's KL divergence spikes at each perilune and returns to zero; the UKF's and the mixture's step up and stay up. 100 tracking trials with a station outage across perilune: the EKF's NEES briefly exceeds the others at perilune; after tracking resumes the UKF is less accurate for a day.

Graduate projects UT Austin

Orbit determination

JahSat-1: orbit determination and smoothing

Estimated a satellite's trajectory from six days of simulated range and range-rate data at three stations (Kwajalein, Diego Garcia, Arecibo), with a 20×20 EGM96 gravity field, Sun and Moon perturbations, box-wing drag and solar radiation pressure. Compared EKF, UKF and batch least squares across data subsets (range-only, range-rate-only, per-station), carried drag as a consider parameter, and applied a fixed-interval unscented smoother, judging each solution by its residuals and by whether its covariance matched the actual RIC error.

EKF / UKFUnscented smootherBatch least squaresManeuver detection

Outcome: the smoother cut epoch position uncertainty from 7.9 m to 29 cm; drag estimates exposed an unannounced maneuver (|ΔV| ≈ 0.4 mm/s) in the day-3/4 data gap; and residuals revealed unmodelled per-station range-rate biases. This project is where my interest in covariance realism started.

Smoothing the JahSat-1 orbit: the unscented smoother cuts epoch position uncertainty from 7.9 m to 29 cm. Day-by-day drag estimates jump 13% at the day-3/4 data gap in every data subset, exposing an unannounced maneuver. Predicted position at the maneuver epoch for each data subset: radial and cross-track agree to about 3 m; in-track is where they differ.

Orbit design & control

Replicating TerraSAR-X orbit control

Recreated TerraSAR-X's 11-day sun-synchronous repeat orbit (167 revolutions per cycle, with the reference closed by differential correction) and its control strategy. Propagated the orbit with J2 and atmospheric drag, sized ΔV burns to keep the ground track within a 250 m tube, and compared burn cadence against an analytic model across the solar cycle.

Orbit propagationSun-sync / repeatDifferential correctionControl budget

Outcome: 13 burns in 66 days (3.1 m/s per year) kept the orbit inside the tube 99.7% of the time; without control, the ground track leaves the tube after 1.9 days.

Keeping TerraSAR-X inside a 250 m tube: 13 burns in 66 days, inside the tube 99.7% of the time. Without control the ground track drifts out of the 250 m tube after 1.9 days; a small burn every ~5 days keeps it inside. Error from the reference orbit at each equator crossing: cross-track driven by drag, radial by eccentricity.

Attitude determination & control

Closed-loop ADCS with an MEKF

Built a closed-loop ADCS simulation for an Earth-pointing satellite: rigid-body dynamics with gravity-gradient, magnetic and drag torques; phase-plane thruster (RCS) control with a ±0.5° dead band; and a six-state multiplicative EKF that fuses gyro, horizon, sun and magnetometer measurements to estimate attitude and gyro bias. Robustness evaluated with a 100-run Monte Carlo over sensor noise, gyro bias and thruster error.

MEKFRCS phase-plane controlSensor fusionMonte Carlo

Outcome: after a ~100 s start-up transient, every Monte Carlo run stays inside the 0.5° dead band; the MEKF finds the gyro bias within about 200 s, and its predicted σ matches the actual error spread over 100 runs, so its error bounds can be trusted.

100 Monte Carlo runs: after a ~100 s start-up transient, every run stays within the 0.5° pointing dead band. The six-state MEKF finds the unknown gyro bias within about 200 s, fusing horizon, sun and magnetometer measurements. Filter consistency: the actual error spread over 100 runs matches the MEKF's own predicted sigma.

Radar remote sensing

SAR image reconstruction

Focused raw ALOS PALSAR L-band echoes over Austin, Texas, with time-domain back-projection. Starting from about 8,400 pulses of 8-bit I/Q samples, I range-compressed each pulse with a Hamming-weighted matched filter, then focused in azimuth and multi-looked a 14.5 km × 2.5 km strip.

SARBack-projectionMatched filteringALOS PALSAR

Outcome: raw echoes with no visible structure became a sharp image in which the river, the street grid and individual bright structures stand out.

Back-projection focused SAR image of Austin, Texas, from raw ALOS PALSAR L-band data. Processing chain: raw echoes, then range compressed, then back-projection focused. Range-compressed data (Hamming-weighted matched filter): each strong target traces a range-migration curve across the pulses.

Earth observation · ML

EO object detection: tree crowns

Benchmarked Faster R-CNN, RetinaNet and SSD for tree-crown detection, integrated DeepForest as a baseline, and prepared a dataset protocol for a fair comparison.

DetectionCNNPyTorch

Outcome: a standardized pipeline for precision/recall benchmarking on EO imagery.

Detection example Training curves

04 · Software

Open-source software

Tools I write to test ideas properly, with validation against reference results and tests that check the claims. Everything is on GitHub.

Navigation & GNSS

Estimation & attitude

Astrodynamics

Simulation & geodesy

05 · Coursework

Graduate coursework

M.S.E. at UT Austin, with a figure or two from each course's work.

UKF RIC error EKF example Computation time

Methods of Orbit Determination

Statistical estimation for trajectory determination.

JahSat-1 OD from range and range-rate at three stations; residuals and RIC error across scenarios. Project →

EKF / UKFCovariance
Profiles at 80°E Spherical-harmonic representation Spherical-harmonic representation

Satellite Geodesy

Gravity-field modeling, functionals and time-variable mass signals.

Degree-variance spectra for potential, geoid, gravity anomaly and ∂²g/∂r²; meridian profiles at 80°E.

Spherical harmonicsGRACE / GRACE-FO
Lagrange points State transition matrix comparison Zero-velocity curves

Celestial Mechanics I

N-body dynamics, perturbation theory, STMs and variational equations.

Extended STM/variational analysis to third-body perturbations; applied Lagrange's planetary equations across regimes.

STMPerturbations
Sun-cycle variation Beta-angle variation Rate of precession

LEO Satellites for Earth Observation

Orbit design for remote sensing: repeat, sun-synchronous and frozen orbits.

TerraSAR-X orbit control around an 11-day repeat cycle; ΔV planning under J2 and drag. Project →

Repeat orbitsOrbit control
Angular velocities Angular velocity components Reaction wheel speeds

Satellite Control Systems

Attitude dynamics, estimation and control.

Closed-loop ADCS with an MEKF, Earth-pointing PD and sensor/actuator error models; Monte Carlo validation. Project →

MEKFADCS
Unfocused vs focused Average spectrum vs reference chirp Multi-look image

Radar Systems for Earth Observation

SAR principles, back-projection focusing, image quality.

Focused SAR image reconstruction via back-projection; multi-look sharpening. Project →

SARImaging
Canopy height Austin Capitol point cloud Slope map

3D Remote Sensing: LiDAR

Point-cloud processing; terrain and vegetation structure.

Tree recognition and canopy-metrics pipeline (DTM/DSM/CHM, crown segmentation) to prototype a KSA workflow.

LiDARCloudCompare
Predicted vs true Cost matrix Validation accuracy

ML Applications in the Geosciences

Object-detection pipelines for EO datasets.

Benchmarked detection architectures for tree crowns with DeepForest as a baseline. Project →

CNNPyTorch

Stochastic Systems, Estimation & Control

Random processes, estimation and stochastic control.

Compared EKF, UKF and the statistically linearized filter on nonlinear systems: consistency, accuracy and observability under varying measurement schedules.

Kalman familySLF

Optimal Control Theory

Pontryagin's maximum principle, dynamic programming, trajectory design.

Low-thrust transfer with indirect shooting, validated against direct transcription; sensitivity to cost weighting and convergence.

Indirect / directTrajectory optimization

Audited: Orbital Debris · Network Control Theory · Statistical Estimation Theory · Applied Orbital Mechanics · Introduction to Remote Sensing

Undergraduate projects UW–Madison

Stress test Damping coefficient Landing legs

Senior Design: Reusable Lunar Lander

NASA Gateway mission concept.

Led shock-absorber design with MATLAB optimization; stability analysis for descent and ascent; CAD and drawings for the vehicle architecture.

Systems engineeringSolidWorks
Model FFT Displacement

Mechanical Vibrations

Fatigue analysis of a solid rocket booster.

Fatigue damage during pad transport via rainflow counting and Miner's rule; proposed an optimal crawler speed and compared with inspection data.

VibrationFatigue
Displacement field Measured strain Analytic strain

Experimental Mechanics

Digital image correlation laboratory.

Measured strain fields; compared homogeneous and inhomogeneous deformation and the limits of optical methods.

DICStrain

06 · Outside work

Beyond engineering, I recharge through nature photography and hiking. Wildlife photography takes the same patience as research: waiting for the right conditions, learning the patterns, and appreciating how the system fits together.

Contact

Let's talk space.

Always happy to talk astrodynamics, spacecraft navigation and control, Earth observation, or open-source space software. Email is the best way to reach me.