Hard problems, composable answers.
Seventeen operational patterns operators build on Nexus Atlas — some pure product capability, some combining Atlas with surrounding systems and procedure, some concepts we’re prepared to build with a partner. Each is labelled. The full operator-depth essays live on the engineering site.
The patterns
CapabilityA fleet as the transport fabric
Aircraft, vehicles or vessels don’t just use the network — they are it.
+
Problem. Every platform depends on its own uplink; lose it, lose the platform.
The pattern. Multi-hop mesh across the formation: spatial redundancy from many platforms, spectral redundancy from many bands, self-healing the moment a lost node is detected (≈1.25 s at default settings).
What Atlas contributes. The entire mechanism — mesh routing, disjoint paths, per-hop encryption. The swarm-fabric deep-dive
CapabilityOne stream, many watchers
More watchers must not mean more copies on the scarcest hop.
+
Problem. Several stations watch the same live feed across the mesh, and per-viewer delivery multiplies the load on exactly the hops with the least to give — starting with the first hop off the aircraft.
The pattern. Declare the feed’s group in the multicast configuration. The mesh computes a distribution tree from the topology it already floods: the shared hops carry one copy however many stations subscribe, and copies are made only at the junctions where the watchers’ paths part. The tree re-forms as the fleet moves; a feed nobody subscribes to costs nothing.
What Atlas contributes. The entire mechanism — subscriber-aware distribution trees with forward-error-correction repair, opt-in and off by default. Measured on an eight-node rig: five minutes of Full-HD-rate video to three stations, with the scarce first hop carrying exactly one copy per packet. How the tree is computed
CapabilityMANET + cellular + VSAT triplex
No single transport survives the whole route.
+
Problem. A vehicle, vessel or aircraft transits mixed terrain — towns with cellular, open country with line-of-sight, dead valleys with neither.
The pattern. Bond a mesh radio, two or more cellular modems on different carriers, and a satellite terminal. The scheduler’s continuous scoring shifts the mix as conditions change — no manual switchover, no flapping.
What Atlas contributes. The bond and the scoring — all shipped.
PatternMulti-constellation satellite bonding
One satellite operator is a commercial and political single point of failure.
+
Problem. A single LEO terminal can be geofenced, throttled or switched off by a decision made far from your mission.
The pattern. Bond terminals from different constellations and a GEO VSAT — different orbits, ground gateways and jurisdictions. When one goes dark, the others absorb the load; a narrowband burst channel carries control telemetry only.
What Atlas contributes. Each terminal is just a link; per-class policy pins control to the resilient path. Terminal procurement and commercial diversity are yours.
PatternHF NVIS — bandwidth of last resort
Everything else is denied; the mission still needs “alive, position X” once a minute.
+
Problem. Every wideband link is jammed, blocked or compromised — but a position report is bytes, not megabits.
The pattern. A near-vertical-incidence HF link covers a wide radius with no infrastructure at kilobit rates. Atlas’s class system routes only critical telemetry through it while the wider links recover.
What Atlas contributes. Tiny-capacity links are first-class citizens — declared capacity, per-class routing, proven down to 64 kbps serial radios. The HF modem integration itself is a pattern, not a shipped adapter.
PatternPre-positioned mesh relays
A 100–300 km dead zone, and the asset can’t carry SATCOM.
+
Problem. The route crosses terrain with no coverage and no line of sight back to base, and the platform is too small for a stabilised terminal.
The pattern. Battery-powered Atlas relay nodes staged along the route — placed, dropped or hosted — wake when the moving asset arrives, hop traffic forward, and sleep again.
What Atlas contributes. Self-forming mesh join, relay routing, store-carry-forward. Power management and emplacement are yours.
PatternCross-border SIMs & in-flight eSIM cycling
Enumerated SIMs get cut at the border.
+
Problem. A long-range platform crosses several jurisdictions; identifiers that sit on carrier rosters too long can be enumerated and deactivated in a coordinated cut.
The pattern. Stage physical SIMs by country; provision eSIM profiles only after departure; push fresh profiles mid-mission from carriers the adversary doesn’t control. Every modem is just another link — the bond holds while profiles cycle underneath.
What Atlas contributes. The bond’s indifference to link identity: endpoint roaming keeps sessions alive through address changes, and a modem that vanishes is just a degraded path. SIM and eSIM logistics are carrier-side systems.
PatternASN & peering diversity
Carriers that share an upstream fail together.
+
Problem. Multiple SIMs from one country often exit through the same upstream network — one routing withdrawal or transit dispute blacks out every modem at once.
The pattern. Curate the SIM stack so each modem exits through a different network path, ideally with at least one egress in a friendly jurisdiction. Correlated cellular outages stop being correlated at the IP layer.
What Atlas contributes. Fully independent per-link measurement and scheduling — the bond treats each exit as its own world. The carrier curation is yours.
CapabilitySneakernet with auto-resume
Hours of RF silence, gigabytes queued.
+
Problem. An asset returns from deep silence with a full buffer, and no link will hold long enough to drain it in one clean pass.
The pattern. Store-carry-forward holds the bulk data — surviving reboots — and drains it at full bonded throughput the moment any path appears: the home network, or a passing relay vehicle, vessel or aerial node.
What Atlas contributes. The entire mechanism — demonstrated as the “data mule” scenario.
PatternDrive-by opportunistic backhaul
Stranded sensors, and something passes within range every few hours.
+
Problem. Fixed sensors, buoys or stranded assets have no economical uplink of their own.
The pattern. A transient vehicle, vessel or aircraft auto-discovers stationary Atlas peers as it enters range, drains their queued telemetry while the geometry holds, delivers pending updates, and moves on. The application never sees the discontinuity.
What Atlas contributes. Dynamic peer discovery, mesh joining, store-carry-forward, roaming — all shipped; the passing platform’s schedule is yours.
PatternHAPS & tethered aerostats as relays
Terrain or policy blocks line-of-sight — and satellite is unavailable.
+
Problem. Ridges, urban canyons or contested airspace separate positions that must talk; satellite is licensed away or politically unavailable.
The pattern. A high-altitude platform — balloon, tethered aerostat or long-endurance fixed-wing — is just another Atlas mesh node, with hundreds of kilometres of radio horizon and its own bonded uplink, so the relay isn’t a single point of failure.
What Atlas contributes. Relay routing and bonding; the platform is yours.
ConceptTime-shifted mission intent
Real-time control is impossible; acting blind is worse.
+
Problem. Heavy interference or a long handover gap breaks real-time control, and the asset can’t wait.
The pattern. Push intent, not commands — “if no contact for 60 s: descend to waypoint, orbit, report”. Delivery rides whatever path touches the asset; autonomy executes; queued telemetry catches the operator up on reconnect.
What Atlas contributes today. The delivery and the queue (store-carry-forward). The intent semantics live in your autopilot or C2 — packaging them as a product feature is a roadmap conversation.
ConceptDPI-resistant transport encapsulation
The only path drops anything not shaped like normal traffic.
+
Problem. A network with deep packet inspection passes only what resembles ordinary web traffic — and volume profile gives away as much as packet shape.
The concept. A pluggable outer wrapper that matches both the shape and the volumetric profile of common protocols. Designed for; not a shipped feature — and stated plainly, because the profile-matching is the genuinely hard part.
ConceptDecoy traffic & volumetric flattening
Even encrypted, mission tempo leaks through volume.
+
Problem. Spikes in link load correlate with operational tempo; an observer of aggregate volume learns timing without breaking any cipher.
The concept. A lowest-priority cover-traffic class keeps every link at a steady envelope — cover shrinks when real traffic appears and refills when it quiets. A natural extension of the class system; designed, not built.
PatternMoving-target ground endpoint
Once your ground station is identified, it’s a target.
+
Problem. The mesh terminates somewhere — and that somewhere invites denial-of-service, route hijack or quiet interception once localised.
The pattern. Run the ground side as a rotating pool of egress points across regions and providers. Peers re-anchor on schedule or on quality drop; sessions survive because authentication is bound to keys, not addresses.
What Atlas contributes. Identity-based sessions, endpoint roaming and relay re-election — the re-anchoring machinery. The egress-pool orchestration is yours.
ConceptPower-budget-aware link selection
On a solar buoy, a satellite bit can cost ten times a mesh bit.
+
Problem. Energy per delivered bit differs wildly across links; naive scheduling drains the battery before the mission window closes.
The concept. An energy hint per link feeding the scheduler — cheap hops for bulk when headroom is low, expensive uplinks reserved for priority traffic. Designed for; today the effect is approximated with per-class policy and capacity declarations.
PatternRecovery seed for electromagnetic events
After an EMP-class event, survivors need a clean baseline.
+
Problem. A high-energy electromagnetic event takes every active radio at once; even surviving nodes may hold corrupted state.
The pattern. One fully-provisioned node kept powered down in shielding boots clean after the event and gossips authoritative configuration to the survivors — a deterministic recovery path for the cost of one spare node and a few kilograms of shielding.
What Atlas contributes. Gossip configuration propagation and mesh join; the doctrine is yours.