SINDY-RUST: A High-Performance Rust Port of PySINDy

[yansu67]

"Hello DynamicsLab team! I'm a big fan of PySINDy. I have built SINDY-RUST, a complete port of the core PySINDy library focusing on extreme performance and Edge AI deployment.

We implemented STLSQ and Polynomial/Fourier/Custom libraries, and successfully identified a complex 10-million sample Noisy Servo Motor dataset in just 5 seconds flat with MODEL ACCURACY: R² = 0.998529

It would be an honor if you could check it out here (https://github.com/yansu67/SINDY-RUST) and perhaps add it to your "Related Projects" section in the README!"

[Jacob-Stevens-Haas]

Very cool. I'd love to learn more about your work/use cases. Seems like a useful example - do you work in industrial engineering?

Also, what are your plans with this repo? Is it a proof-of-concept, do you have large problems that need faster models, or are you looking to develop new methods and write them in rust?

[yansu67]

Hi Jacob! Thanks for reaching out, it's an honor to hear from the DynamicsLab team.

Regarding your questions:

Industrial/Aerospace Focus: Yes, I'm heavily focusing on high-frequency Edge AI and Active Control. A key use case we've implemented is the Kinetix 6500 Servo Motor identification, where SINDy-RS processes 10 million samples across various load regimes (from 'Micro Vibrations' to 'Catastrophic Jams') in just a few seconds to recover the motor's governing electrical and mechanical equations. We also use these discovered models for real-time MPC (Model Predictive Control) and Aerospace Evasion simulations. For these sub-millisecond control loops, the native Rust performance is critical.

Long-term Vision: This project is moving far beyond a POC. My long-term goal for sindy-rs is to build a high-performance framework dedicated to Industrial AI Analytics (Predictive Maintenance) and Edge AI for Military Devices. I aim to enable real-time system monitoring and autonomous decision-making in mission-critical environments where sub-millisecond latency and extreme reliability are non-negotiable.

New Methods: We are actively porting and optimizing advanced methods like HierarchicalSTLSQ and BayesianSINDy to leverage Rust's zero-cost abstractions and multithreading (rayon) for real-time applications.

We'd love for sindy-rs to be a high-performance companion to PySINDy for production and embedded environments! Feel free to check our new unified unified organization/repo at https://github.com/yansu67/SINDY-RS_V2.git