Repository navigation
SINDY-RUST: A High-Performance Rust Port of PySINDy #683
Description
"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!"
Reacted by Jacob Stevens-Haas
Activity
Jacob-Stevens-Haas commentedon Mar 13, 2026
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 commentedon Mar 19, 2026
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