forbes-0023 0aae0f0f94 fix(solver): enforce quaternion continuity on dragged parts during drag (#338)
The _enforce_quat_continuity function previously skipped dragged parts,
assuming the GUI directly controls their placement.  However, Newton
re-solves all free params (including the dragged part's) to satisfy
constraints, and can converge to an equivalent but distinct quaternion
branch.  The C++ validateNewPlacements() then sees a >91 degree rotation
and rejects the step.

Two-level fix:

1. Remove the dragged_ids skip — apply continuity to ALL non-grounded
   bodies, including the dragged part.

2. Add rotation angle check beyond simple hemisphere negation: compute
   the relative quaternion angle using the same formula as the C++
   validator (2*acos(w)).  If it exceeds 91 degrees, reset to the
   previous step's quaternion.  This catches branch jumps where the
   solver finds a geometrically different but constraint-satisfying
   orientation (e.g. Cylindrical + Planar with 180-degree ambiguity).

Verified: all 291 solver tests pass.
2026-02-27 09:37:00 -06:00
2023-10-17 09:56:26 -05:00

Kindred Solver

Assembly constraint solver addon for Kindred Create. Implements the IKCSolver interface via kcsolve.register_solver().

Components

Kindred Solver (Phase 1)

Expression-based Newton-Raphson constraint solver. Pure Python, registered as a Create addon.

  • Expression DAG with eval, symbolic differentiation, simplification: kindred_solver/expr.py
  • Parameter table with fixed/free tracking: kindred_solver/params.py
  • Quaternion rotation as polynomial Expr trees: kindred_solver/quat.py
  • RigidBody entity (7 params: position + unit quaternion): kindred_solver/entities.py
  • Constraint residual generators (Coincident, DistancePointPoint, Fixed): kindred_solver/constraints.py
  • Newton-Raphson solver with symbolic Jacobian: kindred_solver/newton.py
  • Pre-solve passes (substitution, single-equation): kindred_solver/prepass.py
  • DOF counting via Jacobian rank: kindred_solver/dof.py
  • KCSolve IKCSolver bridge: kindred_solver/solver.py

GNN (future phases)

Graph neural network constraint prediction layer. OndselSolver C++ engine and ML training infrastructure. Moved to GNN/ — will be integrated in later phases.

Create Addon Integration

This repo is a git submodule at mods/solver/ in the Create repository. The addon loader discovers package.xml and executes Init.py, which registers the solver:

import kcsolve
from kindred_solver import KindredSolver
kcsolve.register_solver("kindred", KindredSolver)

Testing

python3 -m venv .venv && . .venv/bin/activate
pip install pytest numpy
PYTHONPATH=. pytest tests/ -v

Repository Structure

├── package.xml            # Create addon manifest
├── Init.py                # Solver registration entry point
├── kindred_solver/        # Phase 1: expression-based Newton solver
│   ├── expr.py            # Expression DAG
│   ├── params.py          # Parameter table
│   ├── quat.py            # Quaternion math as Expr trees
│   ├── entities.py        # RigidBody entity
│   ├── constraints.py     # Constraint residual generators
│   ├── newton.py          # Newton-Raphson solver
│   ├── prepass.py         # Pre-solve passes
│   ├── dof.py             # DOF counting
│   └── solver.py          # IKCSolver bridge
├── tests/                 # Unit tests (82 tests)
└── GNN/                   # GNN solver layer (future phases)

License

LGPL-2.1-or-later (see LICENSE)

Description
An assembly solving stack for Kindred Create based on a trained GNN layer between placement actions in the UI and actual constraints applied.
Readme LGPL-2.1 17 MiB
Languages
C++ 66.2%
Python 32.5%
CMake 1.1%