Editing Openai/6944461c-7f90-8000-8c32-54fe2958dd9a (section)

==== ### ====

This Annex defines the requirements for autonomous operation of pressure-driven hydrogen separation systems, including:
* Continuous unattended operation
* Automated fault detection
* Safe, self-directed recovery behaviors

The objective is to ensure systems can operate as infrastructure, not experiments.

===== All systems must demonstrate the ability to operate: =====
* Continuously
* Without manual intervention
* Without real-time human control

for the duration specified in the Competition Guidelines.

Remote monitoring is permitted; remote control is not.

===== Systems must autonomously perform the following functions: =====
* Maintain operating pressure within defined bounds
* Regulate flow rates and separation conditions
* Detect deviations from nominal performance
* Initiate safe responses without external commands

===== Systems must detect and classify faults across at least the following categories: =====

====== - Gradual pressure loss ======
* Sudden pressure excursions
* Seal or containment anomalies

====== - Declining hydrogen output ======
* Purity drift
* Abnormal permeability changes

====== - Membrane damage indicators ======
* Seal compression loss
* Fatigue-related deformation

====== - Sensor drift ======
* Signal dropout
* Conflicting sensor readings

===== Upon detecting a fault, systems must follow a defined response hierarchy: =====
# Stabilization — maintain safe pressure and containment
# Isolation — isolate affected modules where possible
# Recovery — initiate corrective actions
# Degradation — continue operation at reduced capacity if safe
# Safe Shutdown — only if recovery is not possible

Immediate shutdown as a first response is discouraged unless required for safety.

===== Teams must demonstrate at least two autonomous recovery events during testing. =====

Acceptable recovery actions include:
* Flow redistribution
* Pressure rebalancing
* Module isolation and bypass
* Regeneration cycles per Annex C

Recovery must occur without human intervention.

===== Systems must maintain: =====
* Time-stamped fault logs
* Event classification records
* Autonomous decision traces

Logs must be immutable and accessible to judges.

===== Systems must be capable of: =====
* Continuing partial operation after localized failure
* Preventing cascading system-wide faults
* Signaling maintenance needs without emergency shutdown

Catastrophic single-point failures are disqualifying.

===== The following are prohibited: =====
* Scripted or pre-timed fault responses
* Manual overrides during testing
* Remote human decision-making masked as autonomy
* Resetting systems to hide fault history

===== Autonomous control systems must: =====
* Be resilient to signal interruption
* Fail safely under communication loss
* Prevent unauthorized command injection

Cybersecurity theater is not required, but basic integrity is.

===== Judges may: =====
* Introduce controlled faults
* Obscure fault timing
* Disable non-critical sensors

Systems must respond correctly without prior knowledge.

===== XPRIZE retains sole authority to: =====
* Define adequacy of autonomy
* Determine acceptable recovery behavior
* Require additional fault scenarios

All determinations are final.

===== This Annex is binding and applies to: =====
* Annex A (Depth-Based Pressure Compliance)
* Annex B (Submerged Maintainability & Retrieval)
* Annex C (Biofouling, Mineral Scaling & Chemical Stability)
* Annex D (Long-Duration Pressure Creep & Fatigue)

===== Systems demonstrating: =====
* Predictive fault detection
* Self-healing behaviors
* Transparent decision logic

may receive favorable qualitative consideration.

If you want to close the loop completely, the final annexes would be:
* Annex F — Environmental, Cultural & Community Safeguards
* Annex G — Decommissioning, Reuse & End-of-Life Protocols

At this stage, you’re no longer specifying a competition —
you’re specifying how pressure-native civilization-grade systems are allowed to exist.