Open Source Code & Data
The NSF AI Institute in Dynamic Systems develops and shares open-source software, datasets, and benchmarks for data-driven modeling, sensing, optimization, and control of dynamical systems. Explore tools for developing and evaluating AI methods, along with datasets and platforms for applying them to physical systems.
Featured Resources
HydroGym
An open-source reinforcement learning platform for developing and evaluating control methods for fluid dynamics
More on our benchmarks
CTF for Science
A benchmark and evaluation framework for AI methods applied to dynamical systems, providing standardized tasks, datasets, and evaluation tools
Smart Buildings
An open-source building simulation and dataset platform for developing and evaluating learning-based approaches to building energy management and control
Open-Source Software
PyDMD
Dynamic Mode Decomposition for data-driven analysis of dynamical systems
PySINDy
Sparse identification of nonlinear dynamical systems from data
PyKoopman
Data-driven approximation of the Koopman operator
PyNumDiff
Numerical differentiation of noisy data
PySensors
Tools for sparse sensor placement and reconstruction
PyBounds
Tools for analyzing observability along trajectories
DynamicalNodes
Tools for connecting dynamical-system algorithms with live ROS systems
Benchmarks & Datasets
RLE-Bench
A benchmark for evaluating AI agents on robot learning engineering tasks
RoSE Lab Rover Mobility Dataset
A dataset of 1,680 controlled rover mobility trials with synchronized telemetry, motion-capture ground truth, and LiDAR terrain scans
