Adithya Muralikrishna

Work Research

Project

Decision support for space domain awareness

Human-factors research on how operators analyze satellite maneuvers and triage anomalies, and what a decision-support system must do to help them.

Date Aug 2023 – May 2025
By A. Muralikrishna, graduate research assistant, in a multi-university team
Status Research role complete
Filed under Guidance & supervision, Dynamics & astrodynamics

Space domain awareness analysts have to explain what a satellite just did, a planned maneuver, a failure or something deliberate, from sparse tracking data. Within a multi-university research effort, I led the human-factors and decision-making side: what operators need from a decision-support system, written as requirements it can be tested against.

Why it matters for GNC. Autonomy and GNC systems only help if the people supervising them can trust and check their output. This work sits at that boundary, on the same orbit-determination problem a navigation filter solves.

Burn? Not reported Pass 1 Pass 2 Predicted from pass 1 Earth
Figure 1. The decision problem in one picture: a second tracking pass no longer fits the orbit fitted to the first. A planned maneuver, a failure and a deliberate move can all look like this. Illustration, not project data Two-body orbits with an assumed 0.35 km/s prograde burn.

The problem

A tracked satellite rarely announces what it is doing. When a new observation doesn’t fit the orbit fitted to earlier ones, an analyst has to decide what changed and whether it matters: a routine station-keeping burn, a malfunction, or a deliberate move. The data are sparse and noisy, and several explanations can fit the same residuals. Decision-support tools can help, but only if they show operators the evidence in the form their decisions need.

My part

I worked in Georgia Tech’s Cognitive Engineering Center as the human-factors and decision-making lead on a multi-university team; collaborators elsewhere built the dynamics models and ran the Monte Carlo simulations.

  • Requirements: elicited operator needs for maneuver analysis and anomaly triage and turned them into testable system-level functions, then verified the system against them.
  • Behavior and telemetry analysis: modeled and analyzed vehicle behavior and telemetry to check the assumptions the decision-support design rested on.
  • Decision-making: the human-factors side of how analysts reach a call from ambiguous evidence.
Note

Figure 1 is an illustration of the problem class, computed from simple two-body orbits. It is not data or output from the project.

All work