AI-native Hardware R&D

What needed solving.

Two round guides can overconstrain alignment when their spacing is imperfect. The proposed experiment asks whether a round primary guide plus a slotted secondary guide improves planar capture while retaining yaw control, compared with two round guides and a looser two-round alternative.

1. 01Alignment hypothesis
2. 02Parametric fixture + coupons
3. 03Planned physical comparison

Simplified from the implementation and project notes.

How it fits together.

The parametric CAD defines receiver and moving plate, guide pins/bores, connector dummies, axial stops, fit coupons, and biased parts for pitch-error testing. Its constraint model lets the primary round guide locate the plate in plane while the secondary slot releases error along the guide baseline; transverse constraint at that second point still resists yaw. The design is explicitly a planar capture study rather than a six-axis compliant coupling.

Validation software checks BRep overlap in candidate final poses, tolerance-stack translation, final yaw, dummy connector displacement, exported geometry, and stated dimensions. The plan calls for hand approach with free X/Y/yaw motion, staged guide engagement before connector contact, force measurement, mass measurement, and repeated reversible mates across offset and guide-pitch conditions. It gives alternatives and failure observations to record rather than assuming the slotted design wins.

A separate industrial concept remains an envelope/layout study, not a rated robot design. Research documentation argues against extrapolating the tabletop dimensions or force target to commercial connector interfaces, and frames a factory mating/inspection fixture as a hypothesis to investigate.

Later A1 and A2 work adds an OpenCV ArUco vision loop, bounded motion requests, force-sensor fault handling, and an integrated bench-mechanism candidate. The inspection report collects the assembly, engineering files, and validation records. The recorded 210 A1 trials use synthetic camera inputs and an analytic contact surrogate; they are not physical mating trials. A2 adds CAD and hardware-facing integration artifacts while fit, calibration, and stop-loop qualification remain open.

Building and checking it.

The next experiment is to print fit coupons, assemble the unpowered fixture, and compare two-round versus round-and-slot guides under controlled offsets and pitch error. Record capture success, jamming, final alignment, rocking, peak force, and reversibility; only then decide whether motorization would answer a meaningful remaining question.

The current checked workflow regenerates STEP/STL/PNG from source parameters and performs software geometry/export checks. Those checks prepare the physical protocol but do not substitute for its measurements.