```html
Seastead Convoy Mode Design
Seastead Convoy Mode — Design Notes
These notes flesh out a practical "convoy mode" for a fleet of triangular trimaran-foil seasteads traveling together in a loose grid, using shared sensing, mesh networking, and moving-base RTK GPS.
1. What "Convoy Mode" Does
- Station keeping on a grid. Each seastead is assigned a grid cell (e.g., 150 m spacing). Autopilot holds position relative to neighbors — not relative to earth — using moving-base RTK.
- Join / leave protocol. Joining craft approach from outside, reach an open cell, and on passing a threshold (~½ grid spacing with low relative velocity) the system announces "convoy mode engaged."
- Shared situational awareness. All seasteads pool radar, AIS, camera, and AI watch data into a shared track database.
- Distributed night watch. Crews that are "on watch" check in periodically (dead-man switch + positive acknowledgment). The system knows at all times which humans + which AIs are watching.
- Coordinated maneuvering. If the convoy turns, accelerates, or opens up to let weather through, each seastead gets a synchronized trajectory.
- Collision avoidance. Tracked external targets are evaluated against the convoy envelope; if a CPA is threatening, the convoy maneuvers as a unit or opens a gap.
2. Grid Geometry & Spacing
A seastead footprint is roughly an 80 ft × 40 ft triangle (~25 m × 12 m). Reasonable grid spacings:
| Spacing | Pros | Cons |
|---|
| 50 m | Tight; easy comms; close social feel | Wake interaction, less reaction time, hard in swell |
| 100–150 m (recommended) | Comfortable margin; good Wi-Fi range; minutes of reaction time | Slightly weaker mesh SNR |
| 300 m+ | Very safe; less wake interference | Need higher-gain antennas or sub-GHz fallback; harder to see neighbors visually at night |
3. Position Reference: Moving-Base RTK
- One seastead (the "flagship" or elected leader) acts as the moving base. Others compute relative position/heading to it at ~2–5 cm accuracy.
- Hardware: dual-antenna u-blox ZED-F9P or similar (~$200–400 per receiver; two per seastead for heading). Total ~$500–800/seastead.
- RTCM correction stream is tiny (~2–5 kbit/s) — easy to multicast over the mesh.
- Each seastead also keeps an absolute position (standard GNSS + PPP) for AIS and navigation.
- Failover: if the flagship drops out, a pre-ranked successor takes over; relative geometry is unaffected because all receivers are symmetric.
4. Local Mesh Communications
4.1 Why Wi-Fi 5/6 at 5 GHz is the right primary link
- Cheap, commodity, well-supported driver/mesh stacks (802.11s, batman-adv, Meshtastic-for-control-plane, etc.).
- 5 GHz has many non-DFS channels and wide 40/80 MHz options — plenty of spectrum over open water where there is zero interference.
- Directional panels give huge link margin. Over salt water with a clean Fresnel zone, 5 GHz point-to-point at 150 m is trivial — the link budget is the same that WISPs use for 5–15 km shots on land.
4.2 Recommended radio hardware
| Device | Role | Approx. Price | Notes |
|---|
| Mikrotik LHG XL 5 AC / LHG 5 AC | Directional 5 GHz, ~24–27 dBi, to each of 4 neighbors | $90–140 each | Rugged, PoE, 10+ km line-of-sight capable; overkill for 150 m = enormous margin |
| Ubiquiti LiteBeam 5AC Gen 2 | Same role, alternative vendor | ~$100 | Well-known airMAX ecosystem |
| Mikrotik mANTBox 19s / sector | 90° sector (covers front-left & front-right with one radio) | ~$180 | Fewer radios, slightly less gain |
| Omni 5 GHz AP (e.g., UniFi AP-AC-Mesh-Pro) | Short-range backup + "drop-in" for newcomers | ~$180 | For joining seasteads before they're aligned to the grid |
| 900 MHz LoRa / sub-GHz fallback (RAK, Meshtastic) | Low-bandwidth resilient command/telemetry | $30–80 | Works past the horizon; keeps heartbeats alive in fog/rain/outage |
Recommendation per seastead:
- 4× directional 5 GHz radios on the triangle — one facing each cardinal grid direction (N/E/S/W relative to the convoy, not earth).
- 1× 2.4/5 GHz omni (for new joiners, dinghy, phones, and short-range redundancy).
- 1× 900 MHz LoRa node as a last-ditch heartbeat and "I'm still here" beacon.
- Budget: ~$600–900 in radios + ~$150 in cabling/PoE + ~$200 router/switch ⇒ ~$1,000–1,300 total.
4.3 Expected performance at 150 m over salt water
| Metric | Typical |
|---|
| Link rate (80 MHz, 2x2 MIMO, 802.11ac) | 400–866 Mbit/s PHY, ~250–500 Mbit/s TCP |
| Latency | 1–3 ms |
| Link margin | 30+ dB (rain fade & spray are easily absorbed) |
| Range limit before degradation | Several km line-of-sight; sea-state multipath is the real limit, not free-space loss |
Sea-surface multipath is the one non-obvious issue: reflections off the water can cause nulls that move with swell. Mitigations: mount antennas high (7 ft frame + roof gives you ~10–12 ft), use vertical polarization, and prefer 5 GHz over 2.4 GHz because shorter wavelengths fade faster into incoherent scatter.
4.4 Mesh / routing software
- Routing:
batman-adv (layer-2, Linux kernel, very mature) or OLSRv2. Babel is another solid option. Avoid plain 802.11s for more than a handful of nodes.
- OS: OpenWrt on the radios; Debian/Ubuntu on the main seastead computer.
- Data bus: MQTT or NATS for telemetry/tracks; ROS 2 (DDS) if you want tight autopilot-grade pub/sub.
- Time sync: PTP (IEEE-1588) over the mesh, disciplined by GNSS PPS — sub-microsecond, needed for camera parallax and sensor fusion.
- Security: WPA3-SAE on the radios, plus a WireGuard overlay between seasteads. Every message signed with the sending seastead's key. New joiners get provisional read-only access until admitted.
5. Shared Track Database
- Every seastead publishes its own detections (camera, radar, AIS, ADS-B if you have it) to a shared topic.
- A distributed fusion service correlates detections across seasteads. Because camera positions are known to ~2 cm via RTK and time-synced to µs, parallax triangulation from two seasteads 150 m apart gives range accuracy of a few percent out to several km.
- The database is CRDT-based (e.g., Automerge or a custom last-writer-wins map per track ID) so it survives partitions and rejoins cleanly.
- Each track carries: ID, position, velocity, covariance, classification, contributing sensors, last-seen, and confidence.
6. Night Watch & Human-in-the-Loop
- On-watch declaration: a crew member presses "I'm on watch" in the app; they must tap an attention check every ~5–10 minutes (simple, like a truck dead-man).
- Convoy-wide roster: UI shows how many humans and which AI services are currently watching, across all seasteads.
- Minimum watch policy: convoy software refuses to enter "low-alert" mode unless at least N humans + all AI watchers are confirmed active. Otherwise everyone's alarms stay loud.
- Escalation: tracked contact with CPA < threshold → wake the on-call captain on that seastead AND the convoy leader; loud alarms; auto-maneuver proposal displayed.
7. Join / Leave Protocol
- Approaching seastead connects via omni Wi-Fi or LoRa and authenticates (pre-shared key or signed cert).
- Convoy software proposes a grid cell (usually on the edge, downwind or downstream).
- Pilot maneuvers manually toward the cell. UI shows a target "box" in AR / chart view.
- When relative position is within ½ grid spacing, relative velocity < 0.3 kn, and heading aligned within 10°, autopilot offers "Engage convoy mode".
- On engage: directional antennas auto-align to the four new neighbors, RTK moving-base link established, track DB sync begins.
- Leaving is the reverse: click "disengage", autopilot eases thrust to zero relative drift, pilot takes manual control, neighbors reroute the mesh.
8. Convoy-Wide Maneuvers
- Heading change: leader publishes new heading + ramp rate; each seastead rotates the whole grid around the convoy centroid.
- Speed change: simple — everyone's setpoint changes in sync.
- Grid reshape: for weather or to let a ship pass, the convoy can spread (larger spacing) or line up single-file. This is just a new cell assignment broadcast.
- Emergency scatter: every seastead picks a unique bearing that maximizes minimum separation — precomputed and cached.
9. Cost Summary per Seastead (comms + positioning)
| Item | Qty | Unit | Subtotal |
|---|
| 5 GHz directional radio (Mikrotik LHG or UBNT LiteBeam) | 4 | $110 | $440 |
| Omni 2.4/5 GHz AP | 1 | $180 | $180 |
| LoRa / Meshtastic node | 1 | $60 | $60 |
| Managed PoE switch | 1 | $200 | $200 |
| RTK GNSS receivers (dual antenna) | 1 set | $700 | $700 |
| Cabling, PoE injectors, mounts, enclosures | — | — | $250 |
| Edge compute (small x86 or Jetson) | 1 | $400 | $400 |
| Total | | | ~$2,200 |
Starlink is already assumed present, so wide-area backhaul is free in this accounting.
10. Open Questions / Things to Prototype
- Empirically measure the sea-surface multipath nulls at your chosen antenna height and grid spacing — best done with two test rigs on buoys before committing.
- Decide on a leader-election algorithm (Raft is simple and sufficient for < ~50 nodes).
- Define the "convoy cell" coordinate system — probably (bearing, range) from a drifting centroid rather than a fixed earth grid, so the convoy can rotate naturally with wind.
- Legal/COLREGS question: is a formation-keeping group of unmanned-capable vessels one "vessel" for signaling, or many? Lights and shapes need to be thought through.
- How does convoy mode interact with each seastead's three little-airplane stabilizers? The AP setpoint should be the convoy trajectory, not the local wave; the stabilizers just damp pitch/roll as usual.
Bottom line: Four directional 5 GHz Wi-Fi radios per seastead, plus an omni and a LoRa beacon, running batman-adv over WireGuard, disciplined by moving-base RTK and PTP time sync, will give you a rock-solid sub-millisecond, multi-hundred-megabit convoy fabric for about $1k–1.3k of radios per seastead. Everything else — shared track DB, watch roster, join/leave choreography, coordinated maneuvers — is software built on top of that fabric.
```