Open-source research framework
Model how
systems behave.
A Python framework for learning, simulating, and evaluating models of cyber-physical systems.
pip install flowcean
Thermostat / simulation
Modeling approaches
Data-driven and
explicit models.
Data-driven learning
Train models from recorded datasets, incremental streams, or interactions with an active environment.
Learning strategiesHybrid-system modeling
Specify continuous dynamics, event surfaces, and resets, then simulate the resulting hybrid system.
Hybrid systems guidePython example
Simulate a built-in system.
thermostat() creates a benchmark system. simulate() runs it with an input signal and returns a Trace containing the state history, active modes, and transition events.
import numpy as np
from flowcean.hybrid import simulate
from flowcean.hybrid.benchmarks import thermostat
trace = simulate(
thermostat(),
t_span=(0.0, 12.0),
input_stream=lambda _t: np.array([22.0]),
sample_dt=0.05,
)Framework components
Modeling and evaluation.
Environments supply data, transforms prepare it, learning strategies train models, and metrics evaluate their behavior.
Environments & transforms
Load datasets or streams and apply reusable preprocessing steps.
Learning strategies
Train models with offline, incremental, or active learning strategies.
Hybrid simulation
Integrate continuous dynamics and resolve discrete transitions and resets.
Model evaluation
Compare model predictions against reference data.
Research
Project and publications.
Flowcean was developed as part of the AGenC research project on automatic model generation for cyber-physical systems.
AGenC research project- ETFA / 2023IEEE International Conference on Emerging Technologies and Factory Automation
- AIABI at AIxIA / 2024Italian Workshop on Artificial Intelligence and Applications for Business and Industries · ArXiv abs/2603.12015
