Does a tail-risk objective, or a correlated-error model, rescue the angle-stable structural-colour stack?

What it found

Follow-up to garment-stackrobust. The 8-layer stacks are trained on a CVaR objective (mean of the worst quarter of 8 sampled thickness errors) or on errors with a shared rate-drift component (ρ = 0.5), then cross-evaluated on 200 held-out draws under both error models (5% error, 8 seeds, 3000 evaluations). Negative result: no optimised design beats the plain quarter-wave stack (red, independent error: mean cost 14.1 against 16.5 for the best optimised one). CVaR trims Nelder–Mead's 90th-percentile cost (47.1 to 43.1) while raising its mean, and correlated-error training did worse. Stylised thin-film model; nothing was coated or measured.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Code
• Tests

More in Engineered clothing

• How much angle shift can a structural-colour coating shed at a given chroma and oxide mass?Shift against chroma and oxide mass for the garment-stackopt 8-layer coating: on a grid of chroma floors and mass caps, the best red-target shift is 0.68 at a 4.0 g/m² cap but 2.71 at 3.0 and no feasible run at 2.0 or below (floor C* 40); mass binds before chroma. Stylised; 5 seeds per cell, one optimiser.
• Does adding layers cut the angle shift of a structural-colour coating at fixed chroma and oxide mass?Layer-count sweep (2 to 12) of the constrained structural-colour coating: six to ten layers reach best angle shifts under 1 dE at C* 40 and 4 g/m2 where 2 and 4 layers stay near 3.4, but 12 layers did not search well at the same budget (stylised, 8 seeds).
• Does an assistive garment pay for its own mass?One-joint ankle exosuit model: active assist cuts modelled whole-body cost 3–14% (gain 0.1–0.5) but a few kg of garment mass erases it; break-even mass ∝ 1/(mass sensitivity). Stylised; assumed constants, no OpenSim/MyoSuite.
• Is it the motor or the battery that decides whether an exosuit pays for its own mass?Adds a motor and battery mass model to the ankle exosuit: at an assumed 8 N m/kg the motor (about 8 kg) makes a gain-0.3 suit raise cost 0.2–1.9%; the sign flips between 8 and 20 N m/kg, battery hours matter little. Stylised; assumed constants.
• Does it pay to move an exosuit's motor and battery from the shank to the waist?Moves the exosuit's motor and battery from shank to waist: the waist layout wins once a shank kg costs more than 1.03–1.39× a waist kg (assumed ratio, swept); at ratio 3 and 8 N m/kg shank +20.7% vs waist +3.3%. Stylised.
• How much cable bend can a waist-mounted exosuit power train tolerate before the shank layout wins?Replaces the assumed Bowden efficiency with the capstan law exp(-mu*theta): at ratio 3 and mu 0.1 the waist layout beats the shank layout until the cable bends through about 31 rad, so friction loss alone does not decide the layout (hysteresis not modelled). Stylised.

All 45 Engineered clothing projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.