Data Driven MPC Design

Create model predictive controllers directly from experiment data

Data-driven MPC is a control design technique that uses a nonparametric model based on input/output time-domain data to directly (that is without explicit system identification) solve an MPC problem in real-time.

This technique enables you to synthesize an MPC controller using data collected from a single experiment at a nominal operating point. The plant must be LTI and controllable and that the input must be persistently exciting.

For more information, see the corresponding section of What Is Model Predictive Control? and Data-Driven MPC Principles.

Functions

checkPredictionCompare outputs predicted by data-driven model to validation outputs (Since R2026a)
mpcmoveCompute optimal control action and update controller states
simSimulate an MPC controller in closed loop with a linear plant

Objects

DataDrivenMPCData-driven model predictive controller (Since R2026a)
DataDrivenMPCStateMPC controller state (Since R2026a)

Blocks

Data-Driven MPC ControllerSimulate data-driven model predictive controller (Since R2026a)

Topics

Data-Driven MPC Formulation

  • Data-Driven MPC Principles
    Data-driven model predictive controllers use previously collected plant input and output data to compute optimal manipulated variable control moves at each control interval.

Case Studies