Ten system-level conclusions

These ten conclusions hold across the rules, the scoring record, the mass and power model, the interface matrix and the failure record. Each one is stated with the number that drives it and a link to the page where the working is set out. They are ordered by the decision they settle, from the envelope that must be fixed first to the measurement that closes the plan, and a first build can adopt all ten before any detail design starts.

1. The design is set against the frozen envelope, not the 2027 task list

The 50 kg mass limit, the 1.2 m cube and the 70 kg fielded limit have not moved in 8 to 20 years, while the scored content inside them has been rewritten repeatedly. The envelope is therefore the part of the requirement that can be trusted through 2030, and the task list is the part most likely to differ by the time the rover is built S-URC27-RULES S-URC19-RULES §2.a §2.a.ii.

The practical allocation follows: the envelope gets the design margin, and the new content gets the prototype. Mass, volume, power and thermal margin are held against the limits, not against the 2027 objects, and the objects are met with a modular end effector and a modular science bay.

2. The team records the capabilities it will not attempt

2027 is the largest single-year rewrite since the four-mission format settled. It adds a mandatory bioassay, astronaut assistance, 1 m route-finding in hills and a key-and-lock manipulation task in the same rulebook S-URC27-RULES S-URC27-RULES S-URC27-RULES §1.b.v §1.d.ii §1.e. A first build cannot integrate four unproven capability families in the weeks it should be testing one.

A dated list of deferrals, with the scoring cost of each one, is the cheapest risk control in the project. The deferral is a decision recorded in advance, not a capability quietly dropped in March. The scoring cost is taken from the published point values where they exist, and from the mission structure where the 2027 task weights are not yet published: the bioassay and the astronaut sub-mission each sit inside a 100-point mission, and the drone costs 5 kg of the 50 kg limit.

3. The design target is 45–46 kg with 4 kg of reserve

Four of the five champions publish no mass at all, as Mass and power records. WVU won in 2023 at 33 kg S-WVU-SAR2023, and the winning totals in the 200 to 425 band are reached by both the lightest rover and the 42 to 50 kg cluster, so mass does not predict score within the observed range. What mass does is set the failure modes: a rover at 49.5 kg has no room for the last ideas, and the last 3 kg of a first build arrive in the final six weeks.

The reserve is 4 kg at a target of 45 to 46 kg. The 1.2 m cube is the higher-stakes limit of the two, because 40 % of the points is worth the same as 8 kg of overweight at 5 % per kilogram. Keeping the cube is a packaging plan rather than a hope: a folding or removable mast, a removable science module and a full-size mock box that is checked before every configuration change.

4. The pack is sized against the 60-minute Delivery and the back-to-back pair

The duty-cycle arithmetic in Mass and power gives 273 Wh of nameplate for a 60-minute Delivery and 289 Wh for the Autonomy-then-Servicing pair, which is the true single-pack requirement. The recommendation is 400 to 500 Wh of nameplate at 8S Li-ion 21700 in two packs rather than one large one, with 4P good for 180 A comfortably.

The protection chain is ordered, not improvised: controller current limits below the battery management trip, below the fuse curve, below the wire ampacity, below the connector rating. A pack that sagged under peak current would take the arm, the drive and the link down together, which is the failure class the rest of this list is written to avoid.

5. The arm’s mass class is fixed at concept, because it silently chooses the drivetrain

A 4.4, 9.5 or 14 kg arm puts the rover at 30.6, 35.7 or 40.3 kg with the rest of the design held constant. The AK80-9 class of actuator runs out at 40.6 kg, so a single bracket choice can force a +2.7 kg step up in actuator mass. The arm is a mobility decision made in another room.

That is why the arm’s mass class is fixed at concept and carried in the mass allocation from the first week, rather than discovered when the arm is designed. The arm page and the mobility page carry the two sides of the same number.

6. The wheel diameter comes from the box first, then the wheelbase from the terrain, then the motor from the torque

A 250 to 300 mm wheel with a 0.85 to 0.95 m wheelbase and no fold leaves 40 to 50 mm per side for suspension links and tolerances inside the 1.2 m box. A 330 mm wheel already fails the 1.2 m length at a 0.90 m wheelbase S-URC27-RULES §2.a.ii.

The order matters because each step constrains the next. The box fixes the wheel and wheelbase, the terrain fixes the suspension travel and the tread, and only then does the climb torque fix the motor. A rover designed in the reverse order arrives at a motor with ample torque and a chassis that cannot be stowed.

7. The freeze order follows the penalty, not the alphabet

The order of freezes is a risk decision. From most to least expensive to change late: mass and stow, wheel diameter, arm mount, bus voltage, E-stop, spectrum licence, camera count, thermal architecture, mast, autonomy schema, assay pair and configuration plan. The top of that list is a rule breach if it moves after the weigh-in; the bottom is a re-test.

Freezing in that order each time a design review closes keeps the cheap decisions open for as long as possible and the expensive ones closed early. The first 90 days sequence applies the same order to the schedule.

8. The design covers the four load-bearing single points, and the transport leg is rehearsed

Four failures zero several missions at once: wheel retention, the connector in the safety path, pack sag under peak current and the arm’s joint zero. The dependency between missions is measurable: an Equipment zero multiplies the chance of a Delivery zero by 2.9× and an Autonomy zero by 2.1×.

How rovers fail collects the evidence behind those four. The cheapest reliability test is a 30-minute packaged shake test plus a post-arrival checklist, run on the crate that is also the weigh-in box.

9. Comms and autonomy are one reliability problem

Obstructed range should be expected at about one-fifth of line of sight: WVU measured 850 m falling to 300 m in 2023 and 750 m falling to 164 m in 2026 S-WVU-SAR2023 S-WVU-SAR2026. Two bands, a 2 m mast and a directional base antenna are the floor, not a premium option.

Autonomy points are then bought with one simple planner, a progress checker and ten cold-boot full-length runs. The cleanest natural experiment in the record went 0 → 14 → 86 across three seasons while the autonomy stack was simplified, and the reading is that completion rather than sophistication produced the gain S-MONASH-SAR2026. The communications page and the software and autonomy page work the link and the planner as one problem.

10. The first target is the median, with one person owning the coupling numbers

The median finalist has scored between 224 and 258 out of 500 for five years, and lifting each hard mission by 20 points beats any plausible review improvement. The median is the first target because it is the profile of a team whose four missions all work; the winner’s profile adds capability on top of that floor.

Six numbers decide the couplings, and no team in the published record reports all six: arm mass, wheel diameter, mast height, camera count, pack watt-hours and configuration swap time. They should have one owner and one page. One number is worth measuring above the rest: a 30-minute discharge log, which no team publishes.

What would change this list

Four findings would move a conclusion. A 2027 season of data would test the scoring claims in conclusions 2, 3 and 10. A per-task autonomy scoresheet would convert the zero rates into a failure taxonomy. Published masses for the bottom half of the field would turn the mass association into a testable relationship. A team-published reliability metric, such as missions completed without intervention, would replace the zero count as the primary measure of the thing this list is really about.

Further reading

• What separates the top from the rest carries the scoring evidence for conclusions 2, 8 and 10.
• Where the competition is going carries the rule history behind conclusions 1 and 2.
• Mass and power carries the model behind conclusions 3, 4 and 5.
• Interfaces and couplings works the chains that conclusions 5, 6 and 9 depend on.
• How rovers fail carries the failure record behind conclusion 8.
• The first 90 days turns the freeze order into a schedule.
• Mass and power data holds the 118-row table behind the mass conclusions.