Concepts#

These pages explain how mink-warp is put together — the mental model before you dive into solver backends or API details.

Suggested reading order#

Topic

Page

Why this library

Why mink-warp (and not Newton IK)? — purpose vs Newton IK and vs Mink

Hard limits / QP

Constrained IK (hard limits) — box vs general G Δq h ADMM

Layer stack

Architecture overview — Configuration → tasks → solver → integrate

Notation / math

Notation — frames, errors, stacked \(H, c\)

Mink compatibility

Mink API parity — what matches Mink and what differs

Hands-on

Quickstart: single-world IK — single-world IK loop

Scaling out

Batched IKnworld patterns

Demos

Examples — numbered 01_…05_… scripts

Where to go next#

Need

Section

Hard joint / velocity / half-space limits

Constrained IK (hard limits)

Solver choice (DLS / LM / L-BFGS)

Solver backends

Fixed-step control loops

CUDA graphs

Field-level API

API Reference

Throughput numbers

Benchmarks