lighthill 0.0.1

GPU-vectorized per-link hydrodynamics for articulated underwater robots in NVIDIA Isaac Lab

lighthill

GPU-vectorized per-link hydrodynamics for articulated underwater robots in NVIDIA Isaac Lab.

⚠️ Pre-alpha. The package name is reserved on PyPI; the physics engine is in active development. APIs are not yet stable.

What it is

Out of the box, Isaac Lab simulates rigid bodies in air/vacuum. lighthill turns it into an underwater simulator by applying, every physics step and per link:

• buoyancy
• drag
• added mass
• currents

The key difference from existing fast underwater simulators: forces are computed across an articulated robot — vehicle + manipulator(s) — not a single rigid body. That captures the vehicle↔arm hydrodynamic coupling that single-body simulators miss, while staying GPU-vectorized across thousands of parallel environments for modern RL.

• Topology-agnostic: UV, UVMS, multi-arm, swimming-snake — configured, not hardcoded.
• Config-driven: declare links and coefficients; no per-robot force code.
• Validation-first: ships with a suite checked against standard analytical references.

It fills a real gap: fast underwater sims are single-body; multi-body underwater sims are too slow for large-scale RL.

Status

Stage	Pre-alpha (name reservation + scaffold)
Python	≥ 3.10
License	MIT

Name

Named for Sir James Lighthill, whose elongated-body theory of aquatic locomotion is the foundational hydrodynamic model of how slender, articulated bodies generate thrust through reactive (added-mass) forces — exactly the physics this library computes per link.

License

MIT © 2026 Jeff Richley