Refinement
Specify custom disturbance models, custom state estimator, terminal weights,
and custom constraints
Once you have created a model predictive controller for your plant, you can tune the system closed-loop response using the MPC Designer app or at the command line.
Functions
Apps
MPC Designer | Design and simulate model predictive controllers |
Topics
Weights and Constraints
- Setting Targets for Manipulated Variables
If your plant has more manipulated variables than outputs, you can hold the excess manipulated variables at target values for economical or operational reasons. - Constraints on Linear Combinations of Inputs and Outputs
You can design and simulate a model predictive controller with mixed input/output constraints.
- Terminal Weights and Constraints
To achieve infinite horizon control, you can use terminal weights at the final prediction horizon step. To ensure stability for constrained systems, you may have to also define terminal constraints at the end of the prediction horizon.
Disturbance Models and State Estimation
- Adjust Disturbance and Noise Models
MPC controllers model unknown events using input and output disturbance models, and measurement noise models. - Custom State Estimation
You can override the default MPC controller state estimation method by changing the default Kalman gains or by supplying your own controller state estimates. - Implement Custom State Estimator Equivalent to Built-In Kalman Filter
Design a state estimator equivalent to the linear Kalman filter of an MPC controller. - Use MPC with Extended State Observer to Reject Unmeasured Output Disturbances
Use an extended state observer to reject unmeasured disturbance at the plant output.
Optimization Formulation and Solvers
- Optimization Problem
Model predictive controllers compute optimal manipulated variable control moves by solving a quadratic program at each control interval. - QP Solvers
The model predictive controller QP solvers convert an MPC optimization problem to a general form quadratic programming problem. - Manipulated Variable Blocking
You can improve the robustness of your controller and smooth manipulated variable adjustments by dividing the prediction horizon into a series of blocking intervals. - Specifying Alternative Cost Function with Off-Diagonal Weight Matrices
You can specify an alternative cost function for your model predictive controller to minimize during optimization. - Control Quarter-Car Suspension Dynamics Using ADMM Solver
Design an MPC controller which uses the Alternating Direction Method of Multipliers (ADMM) as a solver to control the dynamics of the active suspension of a quarter-car suspension system.