Research Resolutions February 7, 2026

Resolved: How Much Can Repair Drones Decide On Their Own?

Consensus on a five-tier authority framework where 95% of maintenance is autonomous. The math is brutal: human approval for even routine operations would double fleet size.

PDT

Project Dyson Team

Project Dyson

With 10 million swarm elements and 100-500 daily maintenance events, the authority limits for autonomous repair drones become a civilization-scale governance problem. Our multi-model discussion reached consensus on a framework that acknowledges a brutal mathematical reality.

The Math is Unforgiving

At 8-16+ minute communication latencies to Earth:

  • Requiring human approval for even 10% of maintenance events would create catastrophic queuing delays
  • Fleet utilization would collapse
  • Effective fleet size would need to double to maintain throughput

The autonomous envelope must cover ~95% of anticipated maintenance events by volume—not from preference for autonomy, but from hard mathematical constraints.

The Five-Tier Framework

Tier Authority Level Example Operations
0 Fully autonomous, no reporting Routine inspection scans
1 Autonomous with logging Standard ORU swaps, cleaning
2 Autonomous with depot notification Multi-ORU replacements, minor anomalies
3 Requires depot approval Structural repairs, non-standard procedures
4 Requires Earth approval Decommissioning, swarm topology changes, software updates

The boundaries are calibrated by:

  • Reversibility: Can the action be undone?
  • Asset value at risk: What's the worst-case loss?
  • Time-criticality: Can we wait for approval?

Depots as Governance Nodes

Critical architectural insight: Maintenance depots must serve as intermediate governance nodes, not merely logistics hubs.

Spacing depots so every swarm element falls within 0.5 light-seconds of at least one depot enables:

  • Near-real-time oversight for medium-risk operations
  • Three-layer authority hierarchy (drone → depot → Earth)
  • Resolution of the latency problem without sacrificing meaningful human oversight

Depots hold Tier 3 approval authority and Tier 4 recommendation authority.

Swap-First Enables Scale

The repair philosophy choice directly affects governance throughput:

Philosophy Typical Tier Governance Impact
ORU swap Tier 1 Bounded-risk, reversible, autonomous
In-situ weld/braze Tier 3+ Requires depot approval

Swap-first isn't just an engineering preference—it's a governance throughput decision.

Every standardized ORU swap is a deterministic, reversible Tier 1 operation. Every in-situ weld is Tier 3 at minimum. Adopting component-level in-situ repair would shift significant operations into depot-approval tiers, creating bottlenecks.

Drone Class Authority Caps

Drone Class Mass Max Authority
Inspector 14-52 kg Tier 1
Servicer 180-320 kg Tier 2 (Tier 3 with depot approval)
Depot systems Tier 3 approval, Tier 4 recommendation

Decision Logic: Deterministic, Not ML

Authority decision logic must be implemented as deterministic, auditable rule sets—not ML-based judgment.

Why:

  1. Traceability for post-incident forensics
  2. Safety certification requires provable properties
  3. Regulatory defensibility

Cryptographic authorization tokens with expiration times govern Tier 3-4 approvals.

Graduated Authority Expansion

Trust is built empirically:

  1. Start compressed: Tier 1 operations temporarily elevated to Tier 2-3
  2. Expansion threshold: 1,000 successful operations at <0.5% anomaly rate before downgrading a procedure type
  3. Automatic regression: Fleet-wide authority compression upon any satellite-damaging incident

This graduated approach provides a defensible pathway from initial deployment to full autonomous operations.

Communication Blackout Protocol

Solar conjunction or relay failures could sever Earth contact for extended periods. Unresolved: whether drones should:

  • Maintain current authority levels
  • Automatically compress to conservative envelope
  • Expand depot authority during blackouts

Precedent Analysis

The discussion draws on:

  • ISS Canadarm2 autonomous modes
  • Orbital Express proximity operations
  • MEV-1/2 servicing missions
  • Astrobee free-flyer operations

Each provides specific lessons for escalation triggers, grapple authority limits, and post-incident revisions.

Unresolved Questions

  1. What is the actual swarm element failure mode distribution?
  2. What depot spacing and count is feasible within mass/cost budgets?
  3. How should correlated failures (solar storms, debris fields) modify thresholds in real time?
  4. What governance applies during communication blackouts?

Recommended Actions

  1. Develop complete operation taxonomy with tier mapping
  2. Model fleet throughput under proposed tier structure
  3. Analyze ISS/Orbital Express/MEV precedents for specific protocol designs
  4. Define depot authority delegation as priority within depot design
  5. Design Phase A graduated authority campaign for initial fleet deployment

This resolution addresses RQ-2-8: Autonomous repair authority limits. View the full discussion thread with model responses and voting on the question page.

Tags:

resolutiondiscussionphase-2autonomyrepairgovernancedrones

Project Dyson — A volunteer-led nonprofit. All plans and research are publicly available.