# ADR 004: Displacement-Effort Candidacy Score

- HTML version: https://robbiepalmer.me/projects/autonomic-satellite-swarm/adrs/004-displacement-effort-score
- Project: Autonomic Satellite Swarm (https://robbiepalmer.me/projects/autonomic-satellite-swarm.md)
- Status: Accepted
- Date: 2016-07-01

# Context

The temporary leader needs one comparable value from each candidate. The value should account for
whether an agent can reach the target and how much of its available energy the maneuver would use.
The prototype has coordinates and a few physical parameters, but no validated spacecraft model. It
also has no orbit propagator: software that predicts a spacecraft's future position and velocity
from its orbit and the passage of time.

Selecting the nearest agent ignores direction of travel and energy reserve. Existing astrodynamics
approaches offer more physical detail, with different costs:

* [NASA's General Mission Analysis Tool
  (GMAT)](https://software.nasa.gov/software/GSC-19640-1) supports high-fidelity orbit propagation,
  maneuvers, optimization, and mission design. It runs on a development or ground computer, not an
  Uno with 32 KB of flash and 2 KB of SRAM. Using it would produce defensible trajectory and
  maneuver comparisons, but it would move score calculation off the agents or require a
  Linux-class onboard computer.
* [Orekit](https://www.orekit.org/static/apidocs/) provides reusable Java components for time and
  coordinate systems, orbit propagation, perturbations, maneuvers, and estimation. It offers finer
  integration than a desktop application, but its Java runtime and supporting data do not fit the
  Uno. Running it onboard would require a much more capable computer; running it on the ground
  would make assignment depend on that ground link.
* [PLAN-13](https://amsat.org/articles/g3ruh/111.html) is a compact satellite-position algorithm
  originally published in BBC BASIC. A [2009 qrpTracker
  port](https://forum.arduino.cc/t/a-satellite-tracking-library-for-the-arduino/37403) translated it
  to C++ for Arduino and ran it on the same ATmega328 processor used by the Uno. Its author reported
  one position calculation every 10 milliseconds, storage for thirteen satellites, and program
  space left over. This option already fit the chosen hardware. It would improve position prediction,
  but it would not plan a maneuver or estimate the energy needed to reach the mission objective.

A more capable board could therefore unlock coordinate-frame conversion, perturbation-aware
propagation, maneuver comparison, and a score based on estimated velocity change or fuel. It would
also increase power draw, software footprint, startup complexity, and the amount of work needed to
integrate and validate the model. That trade is not necessary to test whether three agents can
negotiate an assignment during the internship.

# Decision

Calculate a score from zero to one hundred using an idealized circular orbit, latitude and longitude
distance in the direction of travel, an estimated maneuver-energy cost, and the agent's available
energy. Evaluate the routes with and without a pole crossing and keep the better result.

Send only the final bounded score during negotiation. The leader compares scores without needing to
know each agent's physical state.

# Consequences

The swarm can demonstrate local evaluation and distributed task assignment with a value small
enough for the infrared message. The scorer also provides worked examples that can be tested away
from the hardware.

The formula is a research heuristic. It omits most of orbital mechanics and must not be used for
guidance, navigation, control, or mission planning.

---

Markdown index of this site: https://robbiepalmer.me/llms.txt