What is SINDy?
Sparse Identification of Nonlinear Dynamical Systems (SINDy) is a relatively new (2016) System Identification technique. It uses an approach called Sparse Regression to find optimal parameters, a key difference to traditional system identification techniques which use linear regression techniques.
Why use SINDy?
1. Convenience
Although SINDy is only a method for conducting system identification, it is different enough from traditional techniques to warrant its own page. Whilst NARMAX is an appropriate technique for nonlinear system identification, it has a key limitation. NARMAX outputs a linear equation in the time domain, whereas SINDy outputs a state space equation, making it particularly convenient to develop models such using advanced control techniques such as Model Predictive Control (MPC).
2. Low Data Resolution
Another significant advantage of SINDy is that it can correctly identify system dynamics even with a small dataset, a significant advantage when compared to NN-based identification techniques.
3. Enforcing Physics
SINDy has the capability to pre-embed physical
4. Handling Nonlinearity
Autoregressive models such as ARX, ARMAX etc. assume that the modelled system is linear.
SINDy for Control (SINDYc)
The SINDy framework was extended in 2018 to encompass control. The paper SINDy for Control demonstrates the potential of SINDy to enable MPC in non-conventional fields.
The SINDYc Framework
