privatepanda.co

papers

Publications

The full write-up of each paper lives here — what the paper does and exactly which parts are mine. Both are in review; preprint and DOI links will be added as they become available.

Under review 02

Multi-Objective Optimization of Computation Offloading in Space Systems under review

Journal of Aerospace Information Systems (AIAA) · submitted Jul 2026 · co-author, 3rd of 4

Where should a computational task run — onboard flight computer, orbital data center, ground station, or terrestrial cloud — when latency, cost, energy, radiation, security, and reliability all conflict? The paper reframes an earlier single-utility model as multi-objective trade-space search: a workload is a split across compute tiers — a point on the tier simplex — and the answer is the Pareto front of feasible splits, not one utility number.

The computational core is my work. The two-stage framework: discrete flight-computer selection under hard TRL and radiation gates, then continuous routing of the workload over the simplex. Both mission evaluators — including the nonlinear one, built from seven separately-ablatable subsystem couplings drawn from spaceflight engineering (M/M/1 queueing on latency, contact-window saturation, link-setup energy, security-protocol overhead, thermal-margin degradation, congestion-driven link availability, silent-data-corruption compounding), all carried into the paper intact. And the six-optimizer benchmark, compared on hypervolume, spacing, and feasible rate, with constraint-domination handling hard mission limits without penalty tuning.

Python · NSGA-II · SPEA2 · MOEA/D · MOPSO · ε-constraint · projected gradient

The Roman Coronagraph Community Participation Program: Target Database and Tools submitted

SPIE Astronomical Telescopes + Instrumentation 2026 · submitted 2026 · co-author

The Roman Space Telescope’s Community Participation Program maintains a target database for the coronagraph instrument — the stars it may observe, with the measurements its planning tools need. The paper describes that database and the tools around it.

My contribution is in the tooling: the importable Python query client — a thin client over the hosted HTTP API, since a database snapshot runs on the order of 2 GB — with server-side star-alias resolution, so any registered catalog name (“HD 95128” or “47 UMa”) returns the same record; Python ports of previously PHP-only server scripts, putting the researcher toolchain in one language; and the all-sky ecliptic-coordinate population, with uncertainties from Monte Carlo draws over a two-piece asymmetric normal that honors the catalog’s separate upper and lower error columns. Merged through two PRs.

Python · SQL · astropy · Monte Carlo