Baseline mode
174.7 Hz FAIL
Problem
What the build needed to solve
My contribution
Built the experiment-to-report pipeline, controller comparisons, mast hand calculations, static/modal FEA, convergence checks, tolerance stack, and preregistered compliance analyzer.
A fast RoboRacer needs defensible vehicle models and a LiDAR structure that does not corrupt the scan plane. The design had to catch failures before fabrication and connect every result to a reproducible run.
Constraints
Design boundaries
First mast mode must exceed the registered 200 Hz guard.
Use the locked 100 mm length, 20 mm OD, and 1.5 mm wall in CAD and fabrication.
Physical compliance needs a 0.001 mm indicator at the planned 20 N load.
Deflection validates compliance, not stress without strain instrumentation.
Build log
Design evolution
Iterations, issues, and fixes documented as a compact notebook timeline.
Revision
Failure mode
Design change
Result
Baseline analysis
The 120 mm × 16 mm mast reached only 174.7 Hz.
Shortened the mast and increased the tube diameter.
The redesigned hand model reached 330.1 Hz.
FEA convergence
A single mesh could make agreement look better than it is.
Ran three refinements and used gauge-region/global metrics instead of singular peak stress.
Final changes stayed below 1.5%; FEA f1 = 285.5 Hz.
Physical gate
A standard 0.01 mm indicator provides only about three counts at 20 N.
Registered a 0.001 mm instrument, fixture subtraction, two axes, five loads, and three cycles.
Executable verdict is ready; measurement remains pending.
Testing
Results
Redesigned FEA
285.5 Hz PASS
Modal guard
200 Hz
Mesh change
<1.5%
MPC p95
1.33 ms
Physical result
Pending
Simulation and FEA evidence is committed and reproducible. The physical test uses five load levels, both axes, at least three load/unload cycles, and fixture-motion subtraction.
VALIDATED requires R² ≥0.99, hysteresis ≤5%, relative U95 ≤10%, and measured as-built compliance within ±15% of as-built FEA.
Scope note: The mast has not yet been fabricated or physically tested. Current structural results are hand calculation and FEA, not measured validation.
Reflection
Engineering lessons
- A failed acceptance criterion is valuable when it causes a cheaper pre-build redesign.
- Convergence and region selection matter more than a visually dramatic peak-stress contour.
- Instrument resolution should be screened against the actual planned load, not the original high-load model point.
Gallery
CAD, fabrication, and test views
CAD
RoboRacer platform context for the modeling and LiDAR-mast design package
Simulation result
Pure-pursuit tracking error (RMS cross-track error) across the lookahead and velocity-gain sweep; the dark basin is the low-error tuning region (simulation).
Links