51 lines
4.6 KiB
Markdown
51 lines
4.6 KiB
Markdown
# Drone command link
|
|
|
|
Proposal, 21 October 2026. This is the radio for Overwatch to command a tasked defense drone. The simulator still uses a generic half-duplex pipe (20 Hz, 250 kbit/s, 50 ms one way) and has not been switched to this waveform.
|
|
|
|
## What the link has to carry
|
|
|
|
One drone is given a line and front-runs it at full speed. The ground side has to send two things: that line, and the latest attacker track so altitude and the line can be rewritten as the radar picture settles. The drone reports its own state on the way back.
|
|
|
|
The bodies are the ones already modeled: a 48-byte command and 64 bytes of telemetry. The approach from line commit to the pass is on the order of two minutes, over a few kilometres of line of sight from a radar site on the 12 km ring to a drone that started on the 9 km loiter ring. The last part of the intercept is the drone's own IR track and does not use this radio.
|
|
|
|
A missed update leaves the drone on the last line it accepted. A forged update must not move that line. The early climb locks a drone onto a heading, so a false track is the failure to design against.
|
|
|
|
## Decision
|
|
|
|
Slotted UHF frequency hopping, half duplex, carrying signed MAVLink 2.
|
|
|
|
The radar site on the inbound side points a directional antenna at the tasked drone. Ten times a second the ground station sends the line and the attacker track. The drone answers with telemetry in its half of the slot. The hop sequence comes from the link key.
|
|
|
|
Two custom MAVLink 2 messages are enough:
|
|
|
|
| Message | Contents |
|
|
|---|---|
|
|
| Line | Position on the line, direction, altitude, time |
|
|
| Track | Attacker position, velocity, time |
|
|
|
|
MAVLink 2 signs the frame. The drone accepts a frame only when the signature checks and the sequence number is newer than the last one it took. Any other frame is dropped. Between accepted frames, and whenever the slots go empty, the drone keeps flying the last line.
|
|
|
|
At an air rate of about 100 kbit/s a command with the existing 24 bytes of framing is 72 bytes, about 6 ms on the air. Telemetry is 88 bytes, about 7 ms. A 10 Hz exchange therefore uses well under a tenth of a second of airtime per second for both directions, and the wait for the next slot is in the same range as the 50 ms delay the simulator already assumes. The drone does not need a stick position at 20 Hz. It needs a newer line.
|
|
|
|
## Interference
|
|
|
|
Each packet occupies one hop and then leaves it. Energy left on a single frequency takes that slot. The next slot is already on another frequency. The directional antenna puts the station's power down the inbound sector rather than across the horizon. Short airtime, a hop set that is not a published channel list, and one aimed path are the resistance this link has.
|
|
|
|
The signature is what stops a rewritten track. Hopping does not. A packet that fails the signature, or that repeats an older sequence, does not become the line.
|
|
|
|
## Channel
|
|
|
|
Fly this inside a coordinated UHF assignment. The hop set is then private, and a 10 Hz exchange is lawful. Ten hertz is several percent of the channel on each side (about 6% for commands and about 7% for telemetry at 100 kbit/s), which a public ISM duty-cycle cap of 1% does not allow.
|
|
|
|
Until that assignment exists, the same two signed messages can run at 1 Hz on an ordinary UHF hopper. Each side then stays under 1% (about 0.6% and 0.7%). The drone still coasts on the last line, and one fix per second is enough to rewrite altitude across a two-minute approach.
|
|
|
|
## Left out of the first build
|
|
|
|
A second band is not part of this proposal. If a trial shows the hop set denied for long enough that the coasted line goes stale before IR locks, copy the same signed frames onto a second radio. The drone takes whichever copy arrives with the newer sequence. The two radios do not have to agree. Requiring both before a command is accepted would let the loss of one band freeze the line.
|
|
|
|
LoRaWAN was set aside for this job. At the spreading factor that reaches a drone out at the loiter ring, one command takes on the order of a second of airtime, and the public channel plan plus the gateway duty cycle forbid a steady stream of them. The processing gain of the chirp helps against weak co-channel noise. It is not a substitute for a moving channel and a signed sequence.
|
|
|
|
## What this does not change
|
|
|
|
The simulator's link settings stay the field-tuning baseline in `simulation/PROJECT_SPEC.md`: 20 Hz slots, 250 kbit/s, 48-byte commands, 64-byte telemetry, 24-byte overhead, 50 ms one-way delay. Those figures describe timing, not this radio. Radar tracks still come from the AERIS sites over the wired path. This proposal is only the path from the command point to the tasked drone.
|