Skip to content
Statistics

Failures and interfaces

The failure record holds 54 sourced entries from 2017 to 2027: 48 with a named mechanism and 6 that record an outcome without a published cause. The interface matrix holds 66 boundaries between twelve subsystems, 37 of them high criticality. Together they describe the same thing from two directions: a failure is an interface that was not managed, and an interface is a place where one subsystem’s change becomes another’s redesign. The failure counts measure how often a mechanism gets published, not how often it happens.

54Sourced failure entries48 classifiable, 6 outcome-only2017–2027
66Interfaces37 high, 25 medium, 4 low2027
9High interfaces on the armCompute and perception also nine2027
2.9×Delivery zero liftGiven an Equipment Servicing zero2022–2026S-URC-RESULTS

Key takeaways

  • Communications is the judges’ named most common field failure, and the record does not contradict them. It ties with the arm at 7 entries, and both sit ahead of drivetrain, software and design process at 6.
  • The most expensive class is autonomy capability, not comms. About 610 of the roughly 945 mission points the field loses to zeros each year are Autonomy points.
  • Zeros cluster into one broken rover. An Equipment Servicing zero multiplies the chance of a Delivery zero by 2.9 and of an Autonomy zero by 2.1.
  • Coupling is concentrated. The arm, compute and perception each carry nine high-criticality interfaces; delivery carries one, because its risk is in the mission sequence rather than a new boundary.

The failure record

Class Entries What breaks
Arm and end-effector 7 Joints, gearboxes, gripper contact, entanglement
Communications and radio 7 Link loss, antenna pointing, video drop
Drivetrain and mechanical 6 Suspension mounts, wheels, belts, auger
Software and autonomy 6 Planners, state machines, late stacks
Design process 6 Not finished, not tested, too much changed at once
Connector, harness and vibration 4 Loose connectors, cable damage, USB loss
Sensor and perception 4 Detection, localisation, camera placement
Operator and human factors 3 Procedure, sequence, command confusion
Logistics and time 2 Mission clock, sequence planning
Thermal 2 Overheating in a sealed box
Power and battery 1 Pack, contactor, distribution
Dust and ingress 0 No team has published a dust-caused mission failure
Outcome only 6 A zero or collapse with no published cause

Of the 54 entries, 51 carry a primary grade and 3 a secondary one. The zero in the dust row is a finding about publication, not about dust: dust is a known hazard with no written post-mortem behind it S-URC-ADVICE.

Failure entries by class and evidence grade
The 54 entries ordered by class, split by evidence grade. Entry counts measure publication frequency, so a class that teams write up, such as the arm, is over-represented relative to one they do not, such as power. Sample: 54 entries from 15 teams plus field-wide statements.Robotics Club @ Mississippi State University, 2026. Original analysis. Full resolution

What the classes cost

Mission Mean score when non-zero Zeros a year Points left to zeros a year
Delivery 27.9 5.0 140
Equipment Servicing 36.3 5.4 196
Autonomy 45.5 13.4 610
Total ≈945

The yearly totals run 994, 1120, 1051, 450 and 1350 for 2022 to 2026. A team that zeroes Equipment Servicing also zeroes Delivery 44 % of the time, 2.9 times the base rate, and Autonomy 85 % of the time, 2.1 times the base rate. That is the signature of one rover-level failure taking several missions down, not of three independent misfortunes.

The exceptions matter. Of the 67 Autonomy zeros, 58 % are isolated: the rover worked everywhere else and could not complete the autonomous task. Only 11 % of Equipment Servicing zeros are isolated. Autonomy is therefore two problems in one column, a capability gap that is the largest persistent loss in the field and a rover-level fault that shows up alongside a mechanical failure.

Share of finalists scoring exactly zero on each mission, 2013–2026
Share of finalists scoring exactly zero on each mission, by year, from 2013 to 2026. Autonomy is the only mission that has never had a zero-free year in the modern era, and it reaches 50 % in 2024. Science has been at zero since 2019. Blank cells are elements that were not in that year’s competition, or 2016, which published no per-task data. Sample: the 2013–2026 official score sheets.Robotics Club @ Mississippi State University, 2026. Original analysis. S-URC13-SCORES S-URC19-SCORES Full resolution

The interface matrix

The twelve subsystems are joined by 66 interfaces, and every pair has at least one. Criticality is applied mechanically: a boundary is high if a change on one side forces a redesign, a re-test or a re-freeze on the other, or if a failure there alone loses at least 10 mission points or breaches a rule.

Feature Count or value
Interfaces 66
High criticality 37
Medium criticality 25
Low criticality 4
Most coupled subsystems Arm 9, compute and perception 9
Next band Environment, systems, communications and mobility at 7 each
Least coupled Delivery at 1
Interface types Mechanical, electrical, data, thermal and timing
The twelve-subsystem interface network and coupling centrality
Panel (a) draws the 37 high-criticality interfaces of the 66, coloured by dominant type, with the 25 medium interfaces dashed. Panel (b) is the coupling centrality, the number of high interfaces each subsystem touches. Sample: the 12 × 12 matrix as written.Robotics Club @ Mississippi State University, 2026. Original analysis. Full resolution

Three features carry most of the warning. The arm and compute are the most coupled subsystems, so deferring either decision is expensive. Every power interface except the thermal one is electrical, and the E-stop rule requires all power draw to cease on one button, which touches power, compute, communications, the arm and the astronaut interface at once S-URC27-RULES §2.a.v. Delivery has a single high interface, the arm’s 5 kg grasp, because it re-uses the existing boundaries rather than adding a new one.

The documented interfaces

The table lists the 41 interfaces that carry a named risk and a documentary trail; the matrix has 66 cells in total, and the remaining ones are recorded in the interfaces analysis.

The values on this page

Statistic Value Definition Period Sample Source Chart and chapter
Failure entries 54 Sourced entries with a mechanism or outcome 2017–2027 15 teams plus field statements Analysis Chart above; How rovers fail
Classifiable entries 48 A named mechanism or symptom 2017–2027 54 entries Analysis How rovers fail
Interfaces 66 Boundaries between twelve subsystems 2027 12 × 12 matrix Analysis Chart above; Interfaces and couplings
High interfaces 37 Redesign-forcing or 10-point boundaries 2027 66 interfaces Analysis Interfaces and couplings
Highest coupling 9 High interfaces on the arm and on compute 2027 66 interfaces Analysis Interfaces and couplings
Zero lift 2.9× Delivery zero given an Equipment zero 2022–2026 162 team-years S-URC-RESULTS Zeros and consistency

Evidence note

The failure counts are counts of publication: a class with seven entries is one that teams write up, not necessarily one that happens most. The judges name communications as the most common cause of a motionless rover but publish no frequency, so the statement is sourced and the number is not. The interface matrix is a structured judgement by the analysts, not a measurement, and the delivery node is the weakest of the twelve because it has no dedicated study. The zero co-occurrence figures are associations across 162 team-years and do not establish which failure caused which.

Sources for this page

  1. S-URC-ADVICEURC Advice (judges' systems-engineering lessons)URC judges and staff.Original Archived copy
  2. S-URC27-RULESUniversity Rover Challenge 2027 – Requirements and GuidelinesThe Mars Society / URC.Original Archived copy
  3. S-URC-RESULTSURC official scores 2022–2026, compiled CSVThe Mars Society / URC, 2022–2026.Print copy