Communications and networking

Communications decides how much of the rover the operators can actually use. Three of the four 2027 missions are teleoperated, and a link that fails does not lose a share of the points, it loses the mission. The rules also make the hard case deliberate: line of sight is not guaranteed, one route target sits behind a hill, and the 900 MHz system has to move between assigned sub-bands. This chapter fixes those limits, compares the candidate bands on computed margins, and recommends a three-plane architecture.

What the subsystem must do

The 2027 rules state the radio problem in one clause: the network must carry control, telemetry and video with no internet, cellular or satellite relay S-URC27-RULES §2.c.i. The course then removes the easy geometry. Line of sight is “not guaranteed for more than 50 %” of the distance S-URC27-RULES §1.c.ii, the second route target is deliberately out of sight S-URC27-RULES §1.e.xvi, and the operators sit inside a metal C2 trailer with no view of the course S-URC27-RULES §3.d.i. A rover that cannot be seen has to be seen through the link.

Four more clauses fix the hardware. The 900 MHz system may occupy at most 8 MHz and must work inside one of three assigned sub-bands, shifting between them per mission S-URC27-RULES §2.c.v. The base antenna is limited to 3 m of height, its base within 5 m of the C2 station, with at least 25 m of cable S-URC27-RULES S-URC27-QA §2.c.iv DQ2. No camera may be used to aim an antenna, which rules out the tracking systems that would otherwise make a dish practical S-URC27-RULES §2.c.iii. Every radio must also be documented as compliant by 2027-04-30, and the judges are explicit that being able to buy equipment is not the same as being allowed to operate it S-URC27-RULES S-URC27-QA §2.c.ii RQ7.

The whole station has to be standing inside the 15-minute set-up window, which makes pre-assembled masts and colour-coded harnesses part of the communications design rather than an operations detail S-URC27-RULES §1.a.ii.

The numbers that size the design

The margins assume compliant EIRP and a stated interference allowance. They are the arithmetic to reproduce before choosing a band; the obstruction, latency and video rows are the ones that decide a mission.

Design space, in one table

At 1 km every option closes, and two effects still kill the link

The headline budgets are comfortable: 900 MHz closes 1 km with 48 dB of margin, 2.4 GHz with 12 dB, 5.8 GHz with 15 dB, all at compliant EIRP. A team that fails at URC did not fail a free-space calculation. Two effects outside that calculation decide the outcome.

Close-in overload. At 5 m the 900 MHz architecture with a 16 dBi yagi puts +2.6 dBm into the receiver, as does 5.8 GHz with the radio at the mast. Most radios specify a −10 dBm maximum input. Every mission begins with a close-in leg, so the design needs transmit power control or a fixed attenuator, tested at 5 m rather than at 1 km.

Terrain. With 20 dB of obstruction the two Wi-Fi bands go negative, at −7.6 dB for 2.4 GHz and −4.9 dB for 5.8 GHz on 25 m of coax, while 900 MHz stays at +28.2 dB. WVU measured the collapse in its own tests: 850 m of line of sight became 300 m obstructed in 2023, and 750 m became 164 m in 2026 S-WVU-SAR2023 S-WVU-SAR2026. The judges’ advice to design for both near and far operation is a direct statement of these two columns S-URC-ADVICE.

Height is worth more than band choice in the open

The first Fresnel radius at the middle of a 1 km path is 9.1 m at 915 MHz, 5.5 m at 2.45 GHz and 3.6 m at 5.8 GHz. With a base antenna at its 3 m limit and a rover antenna at 1.6 m, the line between them is only 2.3 m above the ground at midspan, so the 60 % clearance criterion is met at 5.8 GHz and nowhere else. That is the opposite of the obstruction result, and the two effects pull against each other: low bands need more clearance but lose less when they do not get it.

The practical consequence is that mast height is the cheapest gain on the link. Raising the rover antenna from 0.6 m to 3.0 m recovers about 14 dB at 915 MHz and 14.7 dB at 2.4 GHz on flat ground, which is the same order as switching bands. WVU grew its rover mast from 2 m in 2023 to 3 m in 2025 and reduced antenna mass to pay for it S-WVU-SAR2023 S-WVU-SAR2025. The rover mast is therefore a communications component that the structures and mobility pages also have to carry, and it is the item most likely to break the 1.2 m cube when stowed.

The 900 MHz sub-band is a rules problem before it is a radio problem

The 2027 rule is unusually specific. Occupied bandwidth is capped at 8 MHz, operation must sit exclusively inside 902–910, 911–919 or 920–928 MHz, the schedule assigns the sub-band per mission, and the restriction covers every link between the C2 station, the rover, the drone and the astronaut, including local-area networks between devices at any location, and URC monitors the spectrum on site to police it S-URC27-RULES §2.c.v.

The FCC hop-count rule then depends on the occupied bandwidth, not the channel spacing S-FCC-15-247. Below 250 kHz the requirement is at least 50 hopping frequencies, which an 8 MHz window cannot hold. Between 250 and about 320 kHz the requirement is at least 25, and 250 kHz spacing gives 32, which complies. At 500 kHz the window holds 16, which fails. A team that buys a radio on its data rate and its price has not answered this question.

The concrete trap is in the shipped configuration. The RFD900x-US is a 1 W modem with a published 224 kbit/s ceiling, and it ships with two selectable sub-bands, 902–915 and 915–928, neither of which is one of the three 2027 windows S-VENDOR-RFDESIGN-RFD900X. The workable answer is three saved profiles with the frequency limits set to the assigned window, each verified in set-up. That is a configuration task on every run, and the failure it prevents is a rules breach rather than a performance loss.

Three planes, not one band

WVU is the closest thing to a controlled experiment in the published record: first in 2023, second in 2024 and 2025, fourth in 2026, with a communications paragraph in every System Acceptance Review.

The sequence shows a team moving up in frequency for bandwidth and down in frequency for survival, and ending with three planes rather than a better band choice. The E-stop makes the pattern necessary rather than elegant: the single kill removes all battery power, so the rover radio dies with everything else, and the C2 station sees silence that is expected rather than a fault S-URC27-RULES §2.a.v. A recovery plane that carries a heartbeat and drive commands at low rate is what keeps a bad minute recoverable.

The System Acceptance Review that describes this arrangement is itself the primary source for the 2026 row of the table, and it shows the mast, the antennas and the gimbal camera in service.

Antenna gain belongs at the receiver. Teams routinely put a 1 W radio on a 20 to 29 dBi antenna. That is legal for a fixed point-to-point link in 5.8 GHz and illegal for a moving rover anywhere, because EIRP is capped at 36 dBm S-FCC-15-247. The cap costs nothing in practice: EIRP is what the far end receives, the receiving gain is unregulated, and the correct move is to back the transmitter down and take the gain at the base station. A 29 dBi dish thereby becomes a receive-gain device, which is why the lower-gain sector is the better first choice: the judges warn that the rover will not always be in front of the antenna S-URC-ADVICE.

Interfaces that bite

• Power. The kill removes battery power, so the radio dies with the rover. Link loss after a kill is expected, not a fault, and the field judge has to read the rover’s own status indicator instead. The radio also needs a quiet 12 V rail, because a radio sharing a rail with a motor controller drops packets when the drive moves S-URC27-RULES §2.a.v.
• Sensors and perception. Six 720p30 streams at 3 Mbit/s is 18 Mbit/s, which 900 MHz cannot carry and a contested 2.4 GHz link carries badly. The camera count is a link-budget decision, taken before the cameras are bought.
• Autonomy. The link-loss behaviour is a scoring feature. Autonomy allows a stop-and-return at a 20 % penalty or teleoperation back at 50 %, and Delivery does not allow either S-URC27-RULES §1.e.xviii. The state machine that holds, stops and returns has to work over the recovery plane.
• Operator station. Signal strength, noise floor and latency have to be continuously visible, with a one-action drop to 6 Mbit/s. Without that control the operator cannot trade rate for range.
• Delivery and astronaut. A drone relay is allowed and explicitly encouraged but unproven in this record, and a battery swap counts as an intervention S-URC27-QA S-URC27-QA RQ5 RQ14. A suited astronaut adds a third node on the same bands, and may carry electronics but may not deploy a repeater S-URC27-QA AUQ26.

Failure modes

The failure record is consistent, and it is dominated by physics rather than by component choice.

The gap in that record is worth stating: no public URC post-mortem reports measured link data such as throughput, signal strength or packet loss, so the record is judges’ statements plus each team’s own tested ranges. A team that instruments its link and publishes the numbers would be ahead of the field.

Shortlist for a first build

Open problems

• What is the noise floor at the competition site? There is no measured spectrum data at MDRS. The 3, 6 and 10 dB interference allowances are estimates, and they are the difference between a 12 dB and a 22 dB margin at 1 km. A 60-second scan at a comparable multi-team event settles it.
• What does the 2027 terrain profile actually look like? Base siting and the value of a relay depend on geometry that is not published. A digital elevation model of the area plus the candidate base positions gives Fresnel clearance to the far corners.
• Does the candidate radio shift sub-bands on command? The vendor documents two windows, the rules require three. A bench check with a spectrum display, counting hops inside 8 MHz, is the test.
• What is the receiver’s maximum input? Most radios do not publish a compression point, and the close-in case exceeds the typical limit. A bench test that feeds a known level into each radio finds it.
• What is the real glass-to-glass latency? The 250 ms target is an engineering estimate. A camera pointed at a millisecond clock measures it at 1 km, at 100 m and at three video rates.
• Does a relay earn its place? Two hops turn a failing obstructed link into two comfortable ones, but no team in this record documents a deployment. A two-hop test with a mast at 100 m measures throughput and latency on both legs.

Further reading

• WVU carries four years of published communications detail and the only measured obstruction numbers in the record. Waterloo published its own radio research and is the clearest example of a team documenting the trade rather than the outcome.
• Power, electrical architecture and safety owns the other half of the E-stop problem: the kill that removes battery power also removes the radio.
• Compute, sensors and perception sets the camera count that the link has to carry.
• Interfaces and couplings traces the communications, power and autonomy couplings with numbers.
• The Q&A rulings carries the answers on relays, sub-bands and the base station.