# P01.4 — actual geometric test report

## Release decision: development only

**Do not cast or wear the full-body assembly under load. Independent entry/exit is not demonstrated.** The model has explicitly failed screens and incomplete load-bearing mechanisms. The goal of this pass was to turn the concept into inspectable geometry and expose problems before fabrication.

## What was tested

51 preset-body evaluations: 17 prescribed poses × 3 estimated body variants. An additional 153 single-parameter sweep samples use the nominal body. The entry diagnostic contains 23 straight-translation samples. Six release locations were tested with position-only inverse kinematics. The separate desktop elbow has 136 BREP clearance samples. These counts are **not independent human trials**.

The nominal body is an original rigid-ellipsoid mannequin with 1930.4 mm stature. Shoulder, elbow, hip and other motion is prescribed by mathematical transforms, not solved from a completed exoskeleton's constraint equations. There is no soft tissue, muscle, real scapular motion, migrating knee axis, force feedback, contact pressure, or floor/balance solver. Body weight is not simulated.

## Methods and their limits

**Hard parts against the human:** each mesh triangle is transformed into a body ellipsoid's unit-sphere coordinate system. The minimum point-to-triangle distance from the sphere center is tested. This tests triangle penetration of the modeled ellipsoid at that sampled pose; it is not a surface point-cloud sampling shortcut. A separate screen adds 12 mm to each ellipsoid semi-axis to reserve padding space. Adding to the semi-axes is an approximation, not a true constant-normal-distance offset. The visual human meshes are low-poly renderings of the analytic ellipsoids used in this check.

**Critical hard/hard pairs:** VTK triangle-mesh collision detection, broad-phase bounds, and a contained-vertex check for connected closed solids. Pairs include skin/lames against other skin/lames and against frame/electronics. Joint/pin mating pairs, spring reference volumes, padding contacts, latch targets, and purple unresolved interface volumes are not treated as general hard obstacles. **This is not an exhaustive all-component collision certification.** Coplanar contact and tolerance effects can be flagged as intersections. Mesh tolerances limit the precision; do not interpret a zero count as guaranteed submillimetre clearance.

**Bench joint:** analytic BREP distance between the upper clevis/track and lower tongue/follower at one-degree intervals. This is separate from the whole-body mannequin.

**No continuous collision detection:** clear start/end poses do not imply a clear path. The opening sweep demonstrates precisely that failure.

## Nominal body — preset results

Counts are interfering **part/body or part/part pairs**, not injury severity or millimetres of penetration. Zero in these columns does not mean a usable or safe suit.

| Preset | Hard/body pairs | 12 mm padding-envelope pairs | Critical hard/hard pairs |
|---|---:|---:|---:|
| STANDING | 0 | 0 | 0 |
| ARMS FORWARD | 0 | 0 | 0 |
| ARMS SIDE 90 | 0 | 0 | 0 |
| OVERHEAD 150 | 0 | 0 | 0 |
| ELBOWS 130 | 0 | 0 | 4 |
| REACH FORWARD BENT | 0 | 0 | 2 |
| SEATED 95 | 0 | 0 | 0 |
| HALF SQUAT | 0 | 0 | 2 |
| STEP FORWARD | 0 | 0 | 0 |
| TORSO TWIST 35 | 0 | 1 | 2 |
| TORSO LEAN 30 | 0 | 0 | 2 |
| HEAD TURN 60 | 0 | 0 | 0 |
| HEAD NOD 25 | 0 | 0 | 0 |
| FOREARM ROTATION | 0 | 0 | 0 |
| HIP SIDE STEP | 3 | 6 | 2 |
| OPEN ENTRY LAYOUT | 0 | 0 | 0 |
| DRESSING APPROACH | 0 | 0 | 0 |

The standing preset uses 8-degree arm abduction, not arms pressed tightly against the torso. The open endpoint uses **60-degree shoulder abduction on both arms**. The overhead endpoint uses an idealized fixed shoulder center; actual shoulder translation and the corresponding mechanism remain unresolved. The seated pose has a prescribed pelvis displacement rather than a solved support/contact condition.

## Intermediate motion

| Sweep | Samples | Observed issue span |
|---|---:|---|
| right_elbow | 28 | 130 to 135 degrees (not necessarily every angle inside this interval) |
| right_shoulder_forward | 31 | 10 to 60 degrees (not necessarily every angle inside this interval) |
| right_shoulder_side | 31 | 0 to 5 degrees (not necessarily every angle inside this interval) |
| torso_twist | 15 | -35 to 35 degrees (not necessarily every angle inside this interval) |
| head_turn | 25 | None in sampled angles |
| front_panels_open | 23 | 4.9091 to 68.7273 degrees (not necessarily every angle inside this interval) |

Examples of why endpoint-only checking is insufficient: forward reach at 90 degrees is clear in the nominal screen, but some intermediate forward-raise angles interfere; the fully open endpoint is clear, but parts intersect during the opening path. Full details, exact involved parts, and discrete angles are in `TESTS/sweep_results.json`.

## Main design changes made during this pass

The first layout had 15 nominal standing hard/body intersections and 40 listed hard/hard intersections. Internal cavities, side clearances, frame routing, shell seams, and rear skin extents were revised. The final nominal standing screen has zero in both groups and no intrusion into the approximate 12 mm padding reserve. Earlier logs are retained under `TESTS/iteration_history/`; they describe earlier working geometry, not the current exported files.

A modeled chest ellipsoid plus a 14 mm reserve was used to cut a specific neutral-pose underarm relief from the bicep shell. This is an actual geometry change driven by the human envelope, not an ignored contact. Front elbow and rear-knee compression regions were cut back, the forearm equipment volume moved outboard, and the faceplate opening rotation corrected. **The resulting cosmetic gaps remain unfinished.** They need independently articulated covers, not rigid filling.

## Open issues that block a wearable design

**Opening/escape:** simultaneous opening has intermediate hard/hard conflicts despite a clear open endpoint. Real latches, sequencing, escape releases, stand stability, glove dexterity, release force, and spring unloading are absent. The direct frontward entry translation is only an obstruction diagnostic; it does not model stepping into footwear or independent emergency escape.

**Arms:** the 130-degree elbow preset has four listed hard/hard intersections, including the forearm rail and bicep-front shell. The combined bent-arm reach has chest/bicep interference. Arm adduction below the default 8-degree posture also fails sampled clearances. Retracting/overlapping covers and the actual shoulder mechanism need further design.

**Waist and gait:** torso turning/leaning and the squat produce shell conflicts. In the side-step preset, the outboard thigh rail and thigh skins enter the resting hand's envelope. This is a real layout problem to resolve with rail routing and coordinated arm clearance—not a reason to call the pose safe.

**Fit sensitivity:** the +8% girth variant has four standing underarm hard/body intersections. Therefore the nominal clearances must not be presented as a demonstrated fit for Justin before measurements.

**Structural path:** hip/shoulder/ankle multi-axis mechanisms, lumbar compliance, retained pins, finished connections, harness, stops, and load-rated release system are not designed. The shown frame is not yet a continuous, verified load path to the ground.

**Manufacturing and environment:** casting alloy, wall qualification, pattern draft/parting/cores/shrinkage, inspection, fatigue, thermal management, edge treatment, wiring protection, display fit and sight lines remain unvalidated.

## Gas-spring geometry — rejected direct arrangement

The candidate fixed anchor is 170 mm behind and 120 mm below the elbow in upper-arm coordinates; the moving anchor is 65 mm down the forearm, both outboard by 156 mm. The sampled eye-to-eye distance spans **178.7–273.1 mm**, requiring **94.4 mm** of geometric travel. No commercial spring or force is selected.

The sign of `d(length)/d(angle)` turns negative by the **126-degree** sample, so an extending gas spring reverses its assistance direction near the top of the elbow range. This candidate is therefore **rejected**, not a piston recommendation. The colored anchor blocks and GLB cylinder are reference geometry only, and excluded from full-body clearance approval. A revised lever/cam/cable arrangement needs separate analysis.

## Reach locations — not working releases

All six target positions can be reached by a grip-reference point under the assumed joint limits, but several chosen IK poses intersect the wearer or armor. Grip orientation, finger motion, approach trajectory, operating force, latch travel, and operation with gloves were not solved. These are candidate-point results, not proof that the targets are usable or unusable under every possible posture.

| Target | Hand used | Position residual, mm | Endpoint limb/body pairs | Moving armor/body pairs |
|---|---|---:|---:|---:|
| Q_CHEST_R | L | 0.000 | 0 | 1 |
| Q_CHEST_L | R | 0.001 | 1 | 6 |
| Q_BELT_CENTRE | R | 0.000 | 0 | 6 |
| Q_HELMET_CHIN | R | 0.000 | 0 | 0 |
| Q_FOREARM_R | L | 0.000 | 4 | 10 |
| Q_FOREARM_L | R | 0.000 | 4 | 10 |

## Bench elbow result

The separate desktop clevis/track demonstrator has **0 intersecting samples out of 136**, across 0–135 degrees. Minimum computed upper/lower surface distance is **0.486 mm**. All five parts have valid CAD and watertight mesh exports. This demonstrates only the modeled unloaded joint geometry. It does not transfer that result to the entire suit, the human elbow, printing tolerances, material performance, or spring-assisted operation.

## Evidence

`TESTS/part_manifest.json`, `pose_results.json`, `sweep_results.json`, `entry_translation_diagnostic.json`, `reach_targets.json`, `gas_spring_geometry.json`, `bench_elbow_results.json`, and `package_validation.json`. The source code and parameters are included to reproduce the work.