# ADR 005: Agent Health States for the Apoptosis Demonstration

- HTML version: https://robbiepalmer.me/projects/autonomic-satellite-swarm/adrs/005-agent-health-states
- Project: Autonomic Satellite Swarm (https://robbiepalmer.me/projects/autonomic-satellite-swarm.md)
- Status: Accepted
- Date: 2016-07-01

# Context

The research asks whether a CubeSat can withdraw itself when it can no longer contribute safely to
the swarm. The bench prototype cannot test deorbiting or destruction. It can test whether a node
stops accepting work after local or peer-observed evidence of a fault.

Not every adverse condition should be permanent. Low light may require a temporary power-saving
state, while a fatal local fault or repeated communication failure may require withdrawal that
remains in effect until the board is reset.

# Decision

Represent health with three agent states: alive, asleep, and dead. An asleep agent pauses swarm work
and may return. A dead agent stops updating and cannot rejoin without a reset.

Use a light sensor to demonstrate reversible sleep. If a dependent agent repeatedly sends a score
but hears acknowledgements only for other agents, treat that peer evidence as a failed transmitter
and move the agent to the dead state.

The dead state is the prototype's apoptosis mechanism. It is a software state, not physical damage
or a deorbit command.

# Consequences

The demonstration can distinguish power conservation from permanent withdrawal and can trigger the
latter without destroying hardware. Communication behavior becomes part of health evidence instead
of merely a delivery concern.

The word "dead" is easy to overstate. The evidence thresholds are specific to a three-board demo,
and a real spacecraft would need reset policy, persistence, independent fault evidence, safe-state
actuation, and verification of every irreversible action.

---

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