Open Simulation Examples

Engineering-grade simulation with open-source solvers, set up and run by agents on qBraid compute, and shown in the browser with three.js.

Each example is a complete workflow. It starts from a problem statement, runs a solver on qBraid CPU or GPU instances, checks the result against a known answer, and renders an interactive viewer in the qBraid Agent Canvas. It also shows where quantum computing fits today and where it could fit in the future.

Example	Open-source stack	Stands in for	Benchmark bar (top-10% for the field)	Verdict
wind-tunnel/	OpenFOAM v2412, gmsh	Ansys Fluent, STAR-CCM+	Cylinder Re=100 St/Cd bands; Ahmed body drag vs experiment	Cylinder reached; Ahmed 35° close (0.4% off, mesh not settled); Ahmed 25° not reached (steady-RANS limit)
fleet-routing/	PyVRP, NVIDIA cuOpt, HiGHS, OR-Tools	Gurobi/CPLEX routing, Hexaly	CVRPLIB X: PyVRP's published 0.22% mean gap to best-known	Close: 0.31% at the full budget (30-instance subset); beats the bar below 300 customers
qubit-design/	AWS Palace, gmsh	Ansys HFSS	Modern transmon Hamiltonian; predicted T1 inside IBM medians	Reached (3D-verified: 4.81 GHz, -270 MHz, T1 142–250 µs predicted); Ta-record coherence not reached
drug-discovery/	Boltz-2, RDKit, PoseBusters	Schrödinger Glide, FEP+	Co-folding pose success vs AlphaFold3/Chai; affinity ranking vs FEP+	Poses reached (84% on a filtered subset); affinity close to FEP+ (Spearman 0.77 / 0.87)
topology-optimization/	Python ports of top88/top3d, scikit-sparse	Altair OptiStruct, Abaqus Tosca	Match the published reference codes within 1%	Reached (to ~1e-5%); printable titanium bracket, 88% lighter
robotics/	MuJoCo Playground, Brax, MuJoCo WASM	MATLAB Simulink, proprietary sims	Published Playground Go1 joystick reward and tracking	Reached (reward 28.1 vs ~27; 0 falls); live in-browser policy
energy-grid/	Egret/Pyomo + Ipopt, PyPSA, HiGHS	Siemens PSS/E, PLEXOS	PGLib-OPF best-known objectives within 0.1%	Reached (16 cases to 3,012 buses within 5e-5)
weather/	NVIDIA Earth2Studio, FourCastNet 3	Operational NWP (ECMWF IFS)	WeatherBench2 RMSE vs IFS HRES, days 1–5	Close: FCN3 4-member mean beats HRES on 7 of 9 pairs; day-3 z500 5% worse
finance-risk/	QuantLib, CuPy, Riskfolio-style backtest	Bloomberg PORT, Numerix, MSCI	Basis-point pricing; MC error bars match theory; DeMiguel-protocol backtest	Reached on all three (backtest reproduces "1/N is hard to beat")
materials/	MACE, ORB v3, EquiformerV3, PySCF	VASP, Materials Studio	Matbench Discovery top-10% (F1 ≥ 0.925)	Reached via the #1 MIT-licensed model (F1 0.909 on 250, matches its 0.931 entry); phonons not reached

All verdicts are relative to published references, stated in each README with the date, environment, machine and cost. "Close" means a specific, costed next step would decide it.

Every result in this repo carries a verification stamp: the date, environment, instance profile, wall time and cost.

Layout

• <example>/README.md gives the problem, how to run it, the verified results and the known limits.
• <example>/results/ holds small JSON results that the viewers read.
• <example>/viewer.html is a self-contained three.js viewer.
• skills/ holds draft agent skills distilled from the verified runs. They are being published through the qBraid skills registry (qbraid skills search <topic>); until then, read them here.
• docs/pathway.html is the sector survey these examples came from: open-source alternatives by sector, their licences, and where quantum fits.

Third-party data and licences

The code here is Apache-2.0. Some examples bundle or derive from third-party data, which keeps its own terms:

• Fleet routing: CVRP instances and best-known solutions from CVRPLIB (Uchoa et al. 2017 X set; Christofides/Augerat E and A sets), redistributed for benchmarking with citation. Chicago street network and buildings © OpenStreetMap contributors, ODbL.
• Energy grids: test cases are fetched at run time from PGLib-OPF (CC-BY-4.0); the SciGrid-DE network ships with PyPSA.
• Weather: contains modified Copernicus Climate Change Service information (ERA5), 2026; neither the European Commission nor ECMWF is responsible for any use of it. IFS HRES and ensemble reference scores come from WeatherBench 2. Coastlines from Natural Earth (public domain).
• Drug discovery: protein structures from the RCSB PDB (CC0). Benchmark complexes, and per-complex Vina and GOLD results, from the PoseBusters benchmark (Buttenschoen et al. 2024). FEP+ reference predictions from Schrödinger's public FEP+ benchmark repository (MIT).
• Materials: structures from the Materials Project (CC-BY-4.0) and the WBM set via Matbench Discovery (CC-BY-4.0 data, MIT code). Experimental lattice constants and bulk moduli from Csonka et al. (2009).
• Finance: Ken French Data Library and CBOE index history are downloaded at run time by finance-risk/fetch_data.sh and are not redistributed here.
• Robotics: Unitree Go1 model from MuJoCo Menagerie (BSD-3-Clause).
• Topology optimization: the reference codes top88.m (Andreassen et al. 2011) and top3d.m (Liu and Tovar 2014) are fetched at run time and not redistributed; only a headless driver is included.
• Qubit design: the transmon example and regression reference come from AWS Palace (Apache-2.0).

License

Apache-2.0. The solvers keep their own licences (OpenFOAM GPL-3.0, Palace Apache-2.0, HiGHS MIT, cuOpt Apache-2.0, PyVRP MIT).