LQRegulatorGains
LQRegulatorGains[sspec,wts]
gives the state feedback gains for the system specification sspec that minimizes a cost function with weights wts.
LQRegulatorGains[…,"prop"]
gives the value of the property "prop".
Details and Options
 LQRegulatorGains is also known as linear quadratic regulator, linear quadratic controller or optimal controller.
 LQRegulatorGains is used to compute a regulating controller or tracking controller.
 LQRegulatorGains works by minimizing a quadratic cost function of the states and feedback inputs.
 A regulating controller aims to maintain the system at an equilibrium state despite disturbances pushing it away. Typical examples include maintaining an inverted pendulum in its upright position or maintaining an aircraft in level flight.
 The regulating controller is given by a control law of the form , where is the computed gain matrix.
 The quadratic cost function with weights q, r and p of the states x and feedback inputs u_{f}:

continuoustime system discretetime system  A tracking controller aims to track a reference signal despite disturbances interfering with it. Typical examples include a cruise control system for a car or path tracking for a robot.
 The tracking controller is given by a control law of the form , where is the computed gain matrix for the augmented system that includes the system sys as well as the dynamics for .
 The quadratic cost function with weights q, r and p of the augmented states and feedback inputs u_{f}:

continuoustime system discretetime system  The number of states of the augmented system is given by , where is given by SystemsModelOrder of sys, the order of y_{ref} and the number of signals y_{ref}.
 The choice of weighting matrices results in a tradeoff between performance and control effort, and a good design is arrived at iteratively. Their starting values can be diagonal matrices with entries , where z_{i} is the maximum admissible absolute value of the corresponding x_{i} or u_{i}.
 The weights wts can have the following forms:

{q,r} cost function with no crosscoupling {q,r,p} cost function with crosscoupling matrix p  LQ design works for linear systems as specified by StateSpaceModel:

continuoustime system discretetime system  The resulting feedback gain matrix is then computed from algebraic Riccati equations:

continuoustime system and is the solution to the continuoustime algebraic Riccati equation discretetime system and is the solution to the discretetime algebraic Riccati equation .  The submatrix b_{f} is columns of b corresponding to the feedback inputs u_{f}.
 The system specification sspec is the system sys together with the u_{f}, y_{t} and y_{ref} specifications.
 The system specification sspec can have the following forms:

StateSpaceModel[…] linear control input and linear state AffineStateSpaceModel[…] linear control input and nonlinear state NonlinearStateSpaceModel[…] nonlinear control input and nonlinear state SystemModel[…] general system model <… > detailed system specification given as an Association  The detailed system specification can have the following keys:

"InputModel" sys any one of the models "FeedbackInputs" All the feedback inputs u_{f} "TrackedOutputs" None the tracked outpus y_{t} "TrackedSignal" Automatic the dynamics of y_{ref}  The feedback inputs can have the following forms:

{num_{1},…,num_{n}} numbered inputs num_{i} used by StateSpaceModel, AffineStateSpaceModel and NonlinearStateSpaceModel {name_{1},…,name_{n}} named inputs name_{i} used by SystemModel All uses all inputs  For nonlinear systems such as AffineStateSpaceModel, NonlinearStateSpaceModel and SystemModel, the system will be linearized around its stored operating point.
 LQRegulatorGains[…,"Data"] returns a SystemsModelControllerData object cd that can be used to extract additional properties using the form cd["prop"].
 LQRegulatorGains[…,"prop"] can be used to directly give the value of cd["prop"].
 Possible values for properties "prop" include:

"ClosedLoopPoles" poles of the linearized "ClosedLoopSystem" "ClosedLoopSystem" system csys {"ClosedLoopSystem", cspec} detailed control over the form of the closedloop system "ControllerModel" model cm "Design" type of controller design "DesignModel" model used for the design "FeedbackGains" gain matrix κ or its equivalent "FeedbackGainsModel" model gm or {gm_{1},gm_{2}} "FeedbackInputs" inputs u_{f} of sys used for feedback "InputModel" input model sys "InputsCount" number of inputs u of sys "OpenLoopPoles" poles of "DesignModel" "OutputsCount" number of outputs y of sys "SamplingPeriod" sampling period of sys "StatesCount" number of states x of sys "TrackedOutputs" outputs y_{t} of sys that are tracked  Possible keys for cspec include:

"InputModel" input model in csys "Merge" whether to merge csys "ModelName" name of csys  The diagram of the regulator layout.
 The diagram of the tracker layout.
Examples
open allclose allBasic Examples (7)
The optimal feedback gains for a StateSpaceModel and specified weights:
The optimal feedback gains for a twoinput system:
Use only the first control input:
The feedback gains for a discretetime system:
The feedback gains for a nonlinear system:
The gains have an offset because of the nonzero operating points:
The gains of the approximate linear system do not have the offset:
Scope (32)
Basic Uses (8)
Compute the state feedback gain of a system with equal weighting for the state and input:
Compute the gain for an unstable system:
The gain stabilizes the unstable system:
Compute the state feedback gains for a multiplestate system:
The dimensions of the result correspond to the number of inputs and the system's order:
Compute the gains for a system with 3 states and 2 inputs:
Reverse the weights of the feedback inputs:
Typically, the feedback input with the bigger weight has the smaller norm:
Compute the gains when the cost function contains crosscoupling of the states and feedback inputs:
Choose the feedback inputs for multipleinput systems:
Compute the set of feedback gains for a stabilizable but uncontrollable system:
Compute the gains for a nonlinear system:
The controller is returned as a vector and takes operating points into consideration:
Plant Models (6)
Properties (10)
LQRegulatorGains returns the feedback gains by default:
In general, the feedback is affine in the states:
It is of the form κ.x+λ, where κ and λ are constants:
The systems model of the feedback gains:
An affine systems model of the feedback gains:
The poles of the linearized closedloop system:
Increasing the weight of the states makes the system more stable:
Increasing the weight of the feedback inputs makes the system less stable:
The model used to compute the feedback gains:
The feedback gains structure is different if the design model is directly specified in the input:
Properties related to the input model:
Get the controller data object:
Tracking (5)
The closedloop system tracks the reference signal :
Design a tracking controller for a discretetime system:
The closedloop system tracks the reference signal :
The closedloop system tracks two different reference signals:
Compute the controller effort:
Track a desired reference signal:
The reference signal is of order 2:
Design a controller to track one output of a firstorder system:
ClosedLoop System (3)
Assemble the closedloop system for a nonlinear plant model:
The closedloop system with a linearized model:
Compare the response of the two systems:
Assemble the merged closed loop of a plant with one disturbance and one feedback input:
The unmerged closedloop system:
When merged, it gives the same result as before:
Explicitly specify the merged closedloop system:
Create a closedloop system with a desired name:
The closedloop system has the specified name:
The name can be directly used to specify the closedloop model in other functions:
Applications (12)
Mechanical Systems (2)
The system modeled as an inverted pendulum:
The openloop system is unstable:
Design a balancing controller:
Compute the response of the balanced system to a nonzero initial condition:
Tune the damping of a car suspension:
The response of the suspension to a bump on the road is very pronounced:
Design an optimal controller to dampen the oscillations in the response:
Electromechanical Systems (3)
Without feedback, the ball falls down:
Design an optimal controller to levitate the ball:
The response of the ball with the controller:
The response of the other variables:
Regulate the angular position of a motor:
The openloop response to a torque disturbance:
Design a controller to reject the disturbance:
Regulate a ball and beam system:
If the beam is not horizontal, the ball rolls away:
Design a state feedback controller to return the ball to its original position:
The controlled system returns the ball to its original position:
Aerospace Systems (2)
Improve the handling qualities of an aircraft:
The openloop response to a disturbance in the pitch angle takes about seconds to stabilize:
Compute a set of regulator and estimator gains:
Assemble an estimatorregulator:
The handling response is improved:
Stabilize the longitudinal dynamics of a helicopter:
A model of the helicopter's dynamics:
The response to a change in the forward pitch angle:
The closedloop system is stable:
The helicopter's response has been stabilized by the controller:
Electrical Systems (2)
Tune the response of an RLC circuit:
A descriptor model of the system:
The closedloop systems for each set of weights:
The output response for each set of weights:
The openloop pole locations reveal the system is stable:
However, the system's response has near resonance at frequencies close to 1000 rad/s:
Design a set of controllers with increasing gains to dampen the response:
The closedloop pole locations:
The response is better damped close to 1000 rad/s as the weight is increased:
Chemical Systems (1)
Nautical Systems (1)
Regulate the heading of a boat:
A state space model of the system:
The poles indicate the system is marginally stable:
The heading angle deviates from the equilibrium for nonzero initial conditions:
Design a feedback controller to regulate the boat's heading:
Properties & Relations (24)
The optimal gains are computed for negative feedback:
The closedloop poles with negative feedback :
They are the same as the computed poles:
The closedloop system is obtained using state feedback:
Obtain it directly as a property:
For a nonlinear system, the gains are affine and of the form :
The gains of the linearized system are linear of the form :
The affine gains are obtained by solving :
An LQR design has guaranteed gain and phase margins:
The phase margin is at least :
This is because the Nyquist plot always lies outside the unit circle centered at :
A controllable standard StateSpaceModel can be completely controlled using state feedback:
The closed and openloop poles:
A controllable nonsingular descriptor StateSpaceModel can also be completely controlled:
The closed and openloop poles:
Only a subsystem of an uncontrollable standard StateSpaceModel can be controlled:
The eigenvalue is controllable and is not:
The uncontrollable eigenvalue is not affected:
Only the controllable slow subsystem of a singular descriptor system can be controlled:
The dimension of the slow subsystem:
The gain corresponding to the fast subsystem is :
The pole of the slow subsystem is moved but the pole at of the fast subsystem is unchanged:
The complete system is uncontrollable and is of order less than the number of states:
A singular descriptor system with an uncontrollable slow subsystem is completely uncontrollable:
The dimension of the slow subsystem:
None of the poles are changed:
A stabilizable system can be stabilized using state feedback:
One of the eigenvalues is unstable:
The system is uncontrollable but stabilizable:
LQRegulatorGains can compute the stabilizing gains for the system:
The state feedback brings the unstable eigenvalue to the left halfplane and leaves the other unchanged:
For continuoustime systems, the gains are computed using RiccatiSolve:
LQRegulatorGains gives the same result:
For discretetime systems, the gains are computed using DiscreteRiccatiSolve:
LQRegulatorGains gives the same result:
A larger state weight results in a faster state response:
A larger control weight results in a slower state response:
For a stable system, a larger control weight causes closedloop poles to approach openloop poles:
The distance of the closedloop poles from the openloop poles:
For an unstable system, a larger control weight causes closedloop poles to approach the negative of openloop poles:
The distance of the closedloop poles from the negative of the openloop poles:
LQRegulatorGains and FullInformationOutputRegulator give the same results:
LQRegulatorGains and StateFeedbackGains yield the same results for a singleinput system:
StateFeedbackGains with the closedloop poles of the LQRegulatorGains design:
LQRegulatorGains gives the same gains:
An estimatorregulator is assembled using LQRegulatorGains and EstimatorGains:
Compute the regulator and estimator gains:
Obtain the estimatorregulator using the computed gains:
The optimal cost is a Lyapunov function:
Its Hessian is positive definite:
The state trajectory projected on the optimal cost surface asymptotically approaches the origin:
The optimal cost satisfies the infinite horizon Hamilton–Jacobi–Bellman (HJB) equation:
The optimal costate trajectory:
The optimal input is a extremum:
The solutions of the HJB equation:
The solution is the optimal cost:
The input based on the HJB solution is the computed optimal input:
The costate based on the HJB solution is the computed optimal costate:
The optimal solution satisfies , where is the Hamiltonian:
The optimal solutions of the state, costate and input:
The optimal input minimizes the Hamiltonian, thus satisfying :
State feedback does not alter the inputblocking properties of a system:
The closedloop system for a specific set of weights:
Both the open and closedloop systems block the input Sin[2t]:
The weight that minimizes ρ c.c x(t)^{2}+u(t)^{2}t results in a closedloop system with poles given by the symmetric root locus:
The symmetric root locus plot:
The poles on the plot for the parameter value are those of the closedloop system:
Possible Issues (3)
If has unobservable modes on the imaginary axis, there is no continuoustime solution:
The zero eigenvalue is unobservable:
If has unobservable modes on the unit circle, there is no discretetime solution:
The eigenvalue 1 is unobservable:
The gain computations fail if the control weighting matrix is not positive definite:
Text
Wolfram Research (2010), LQRegulatorGains, Wolfram Language function, https://reference.wolfram.com/language/ref/LQRegulatorGains.html (updated 2021).
CMS
Wolfram Language. 2010. "LQRegulatorGains." Wolfram Language & System Documentation Center. Wolfram Research. Last Modified 2021. https://reference.wolfram.com/language/ref/LQRegulatorGains.html.
APA
Wolfram Language. (2010). LQRegulatorGains. Wolfram Language & System Documentation Center. Retrieved from https://reference.wolfram.com/language/ref/LQRegulatorGains.html