PINN CORE::2341429
37.7749° N·122.4194° W
PHYSICS-INFORMED  DYNAMICS
SYS.STATUS
PINN ENGINE OPERATIONAL[ SF ]--0 FAILURES--0 DELAYS

ASTRAEA / READY

UNIT 01-F / RUNNING

[ HIGH OUTPUT ]

[[[ WARNING ]]]

PINN ENGINE DEPLOYED

MISSION READY>_

ASTRAEA
© 2026
A

// Subroutine 001 — Brief

Bring your governing equations and sparse sensor data. Astraea builds optimized physics-informed neural-network surrogates that drop into your MPC, RL, or state-estimation pipeline — with automated architecture search, loss balancing, and convergence diagnostics.

// Request early access

// Or install the core

$pip install astraeaSee use cases →
DIAG.LOG
SIM-TO-REAL GAP / 4 ALERTS::ALL UNRESOLVED

// The problem

YOUR SIMULATOR'S
PHYSICS ARE WRONG.

Policies trained in simulation fail on real hardware because physics engines use approximated contact, linearized friction, and simplified dynamics. The sim-to-real gap is a physics problem, not a data problem.

ERR.001[[[ WARNING ]]]

Simulators approximate

MuJoCo, Isaac Sim, and PyBullet relax friction cones, soften contacts, and use penalty-based compliance. A policy that hits 95% in sim drops to 30–60% on real hardware.

ERR.002[[[ WARNING ]]]

Data collection is impossible

Underwater, surgical, nuclear, agricultural — you can’t run 10,000 demonstrations when each deployment costs thousands, or when the environment is lethal.

ERR.003[[[ WARNING ]]]

PINNs work but take weeks

Physics-informed neural networks solve this — but manually choosing architectures, loss weights, sampling strategies, and debugging convergence takes weeks of expert tuning.

ERR.004[[[ WARNING ]]]

No tool exists

Every PINN for robotics is a one-off research effort.

PROC.FLOW
EQUATIONS IN / DYNAMICS MODEL OUT::4 STAGES

// How it works

EQUATIONS IN.
MODEL OUT.

Your governing equations are the input, not the code. Astraea validates the problem, searches architectures and losses automatically, and emits a surrogate ready for your control pipeline.

STAGE / 01

DEFINE

Provide your ODEs/PDEs — Fossen, Cosserat, Navier-Stokes, or any custom equations. Add boundary and initial conditions.

STAGE / 02

VALIDATE

Automatic well-posedness checking. Classifies equation type. Flags stiffness, discontinuities, and ill-conditioning before wasting GPU hours.

STAGE / 03

SEARCH

Bayesian optimization over architectures, activation functions, loss weights, and sampling strategies. Self-adaptive training balances residuals automatically.

STAGE / 04

DEPLOY

Export to CasADi, Drake, ROS, ONNX, or TorchScript. Real-time inference at 100+ Hz on GPU. Full diagnostics and error bounds included.

OUTPUT
CASADI·DRAKE·ROS 2·ONNX·TORCHSCRIPT::100+ HZ ON GPU
MSN.GRID
ONE ENGINE / MANY PHYSICS::06 ACTIVE PROFILES

// Use cases

ONE ENGINE.
MANY PHYSICS.

Every use case has the same structure: known governing equations, sparse real-world data, need for an accurate dynamics model. The equations are the variable — the pipeline is the same.

MSN-01/ ACTIVE

Underwater Robotics

Mν̇ + C(ν)ν + D(ν)ν + g(η) = τ

Fossen 6-DOF dynamics with ocean current estimation. Correct DVL drift, estimate current fields, plan paths through currents from sparse IMU and DVL data.

FOSSEN 6-DOFCURRENT ESTNAV CORR
MSN-02/ ACTIVE

Soft Robotics

ρA ∂²r/∂t² = ∂n/∂s + f

Cosserat rod and continuum mechanics for pneumatic actuators, cable-driven arms, and deformable manipulators. Proven 44,000× speedup enabling MPC at 70 Hz.

COSSERATCONTINUUMMPC 70HZ
MSN-03/ ACTIVE

Surgical Robotics

σ = C : ε + η ∂ε/∂t

Deformable tissue modeling for surgical planning and force estimation. Every anatomy is different — PINNs generalize from sparse force-torque data where simulators fail.

VISCOELASTICTISSUEF/T EST
MSN-04/ ACTIVE

Industrial Manipulation

M(q)q̈ + C(q,q̇)q̇ + τ_f(q̇) = τ

Friction-inclusive robot dynamics for high-precision industrial arms. Physics-informed friction models that transfer across operating conditions.

RIGID BODYFRICTIONCONTACT
MSN-05/ ACTIVE

Hazardous Environments

F = ma + F_contact + F_constraint

Nuclear decommissioning, deep mining, space exploration. When data collection is dangerous or lethal, physics constraints replace the demonstrations you can’t collect.

CONTACTREMOTE OPSSAFETY CRIT
MSN-06/ ACTIVE

Agriculture / Deformable

∇·σ + ρb = ρü

Soft contact with plants, fruit, and soil. Variable terrain, seasonal changes, deformable grasping. Each object is unique — physics generalizes where data memorizes.

DEFORMABLEFLUID-STRSOFT GRASP
OPS.PARAMS
EQUATION LIBRARY / VALIDATED TEMPLATES::08 INSTALLED

// Equation library

BUILT-IN TEMPLATES.
OR BRING YOUR OWN.

Start from validated equation templates with known-good architecture configs, or define any custom ODE/PDE system. Each template ships with tested hyperparameter ranges.

01 / DYNAMICS
Fossen 6-DOF
Mν̇ + C(ν)ν + D(ν)ν + g(η) = τ
02 / STRUCTURAL
Cosserat Rod
∂n/∂s + f = ρA · ∂²r/∂t²
03 / FLUIDS
Navier–Stokes 2D/3D
∂u/∂t + (u·∇)u = −∇p/ρ + ν∇²u
04 / TRANSPORT
Advection–Diffusion
∂c/∂t + u·∇c = D∇²c + S
05 / THERMAL
Heat Equation
∂T/∂t = α∇²T + Q/(ρcₚ)
06 / MECHANICS
Linear Elasticity
∇·σ + f = ρü
07 / REACTION
Reaction–Diffusion
∂u/∂t = D∇²u + R(u)
08 / CUSTOM
Any ODE / PDE System
F(u, ∂u/∂t, ∂u/∂x, …) = 0
DEPLOY
CORE / PRO TIERS::OSS · APACHE 2.0

// Open source + commercial

THE CORE IS FREE.
SCALE WHEN READY.

The PINN engine and equation library are open source under Apache 2.0. AutoML, managed compute, and deployment integrations are commercial.

STANDARD ISSUE / APACHE 2.0[ FREE ]

ASTRAEA CORE

Full PINN engine, equation library, and training loop. Everything you need to build physics-informed dynamics models.

  • [+]Complete PINN training engine
  • [+]Built-in equation templates
  • [+]Self-adaptive loss weighting
  • [+]Sensor ingestion (CSV / ROS / HDF5)
  • [+]Solution + residual visualization
  • [+]ONNX + TorchScript export
  • [+]Community equation contributions
  • [+]Full docs + tutorials
PRO-GRADE / COMMERCIAL[ HIGH OUTPUT ]

ASTRAEA PRO

Automated architecture search, managed GPU training, and production deployment integrations for teams shipping real robots.

  • [⚡]AutoML arch + loss search (Bayesian opt)
  • [⚡]Well-posedness validation pre-train
  • [⚡]Managed GPU training (A100 / H100)
  • [⚡]CasADi, Drake, ROS 2 export
  • [⚡]Convergence guarantees + error bounds
  • [⚡]Validated hyperparams per domain
  • [⚡]Model versioning + deploy tracking
  • [⚡]Priority support + consulting
DIAG.CMP
LLM vs PINN LIB vs ASTRAEA::08 METRICS

// Why a tool?

YES, YOU COULD
PROMPT AN LLM.

There's a difference between generating code and reliably producing dynamics models that control real robots.

CAPABILITYLLM + AGENTPINN LIBASTRAEA
Generate PINN code✓ with bugs✓ manual✓ automatic
Architecture searchvibes-basedmanualBayesian opt
Loss weight tuningtrial / errormanualself-adaptive
Well-posedness checknonoautomatic
Convergence guaranteenonovalidated configs
MPC / RL pipeline exportmanual wiringmanual wiringCasADi · Drake · ROS
Time to working modeldays–weeksweekshours
Reproduciblenopartiallyfully versioned

[[[ MISSION READY ]]]

YOUR EQUATIONS
KNOW MORE.

Astraea is in early development. Star the repo, join the Discord, or reach out if you're working on physics-informed dynamics for robotics.

// Or request Astraea Pro early access