clj-physics

A Data-Oriented Toolkit for Multi-Domain Surrogate Modeling & Simulation

clj-physics is a high-assurance, idiomatic Clojure library designed for modeling, simulation, and synthetic data generation.

Unlike traditional finite-element solvers that prioritize microscopic fidelity at the cost of computational expense, clj-physics focuses on surrogate modeling, fast-time simulation, and uncertainty quantification. It provides a unified data-driven interface for rigid-body dynamics (6-DOF), computational fluid dynamics (CFD) approximations, and electromagnetic (EM) propagation.

Philosophy

The library is built on three core principles:

1. Data as Physics: All physical entities—from airframes and material properties to flow fields and RF signals—are represented as immutable, schema-validated maps. This facilitates transparent introspection and easier integration with machine learning pipelines.
2. Surrogate & Correct: For complex phenomena like fluid flow, the library employs a "Surrogate + Corrector" pipeline. It generates fast initial estimates using analytical models (or hooks for ML inference) and then enforces physical plausibility (mass conservation) using a divergence-free projection solver.
3. Numerical Robustness: The system prioritizes stability over raw speed. Critical arithmetic operations, particularly in electromagnetics, are guarded against underflow/overflow and ill-conditioned states, falling back to BigDecimal precision when necessary.

Key Capabilities

🌊 Surrogate CFD

Avoids the computational cost of full Navier-Stokes solvers by using reduced-order models suitable for mission planning and real-time synthesis.

• Helmholtz-Hodge Decomposition: Implements a grid-based Poisson solver (Conjugate Gradient method) to project analytical flow fields (like Gaussian plumes) onto a divergence-free manifold, ensuring mass conservation and boundary enforcement.
• Domain Wrappers: Specialized APIs for Atmospheric Plumes (dispersion modeling), Maritime/Shallow-Water (wave fields), and Debris Transport.
• ML Integration: A registry system allows the injection of external machine learning models (e.g., ONNX) to serve as the surrogate prior.

📡 Electromagnetics

A robust engine for modeling RF propagation, safety envelopes, and material interaction.

• Stochastic Propagation: Built-in Monte Carlo engine for propagating plane waves through media with statistical uncertainty.
• Phasor Arithmetic: Full support for harmonic fields ($e^{j\omega t}$ convention), superposition, and polarization.
• Safety Constraints: Automated evaluation of power density ($W/m^2$) and field amplitude limits.

✈️ Dynamics & Environment

A complete 6-Degrees-of-Freedom (6-DOF) simulation stack.

• Integrators: Adaptive Runge-Kutta-Fehlberg (RKF45) and classic RK4 integrators.
• Platform Models: Validated models for Fixed-Wing Aircraft, UGVs, USVs, and Submarines.
• Environment: High-fidelity WGS84 coordinate transforms, 1976 US Standard Atmosphere (up to 86km), and UNESCO 1983 Ocean Equation of State.

Install

Pull the latest version from Clojars:

;; deps.edn
net.clojars.helsingin/physics {:mvn/version "RELEASE"}

Quick Start

(require '[physics.electromagnetics.fields :as f]
         '[physics.electromagnetics.materials :as m]
         '[physics.electromagnetics.propagation :as p]
         '[physics.frames :as frames]
         '[physics.environment :as env]
         '[physics.dynamics :as dyn])

;; 1. Build a field and propagate it
(def field (f/->field {:type :electric :frequency-hz 2.4e9 :amplitude 1.0 :orientation [0 0 1]}))
(def air (m/->material {:name "air" :type :dielectric :permittivity-rel 1.0006 :conductivity-s-m 0.0}))
(p/propagate-plane-wave field air 100.0)

;; 2. Frame transforms (WGS84 <-> ECEF)
(def origin {:position [37.62 -122.38 0.0]})
(frames/geodetic->ecef {:lat-deg 37.62 :lon-deg -122.38 :alt-m 0.0})

;; 3. Dynamics: compute derivatives for a fixed-wing state
(def model (physics.models.common/fixed-wing))
(def state {:position [0 0 1000]
            :velocity [60 0 0]
            :orientation (physics.core/euler->quaternion {:roll 0 :pitch 0 :yaw 0})
            :angular-rate [0 0 0]})
(dyn/rigid-body-derivatives model state (env/isa-profile 1000) {:throttle 0.5})

Tests

clojure -M:test

Tests cover core math, frames, environment, dynamics, constraints, integrators, observer, CFD, electromagnetics, and spatial utilities. A default tests.edn is included for Kaocha with a pinned seed.

Build and release

• Build the jar: clojure -T:build jar
• Deploy to Clojars: make deploy (reads credentials from ~/.m2/settings.xml server clojars, or falls back to CLOJARS_USERNAME/CLOJARS_PASSWORD if exported)

License

Licensed under GPL-3.0-only. See LICENSE for details.