functionValue

Method of mss.

Evaluate the state equation at a given state and input.

Source: src/model/@mss/functionValue.m

Syntax

f = functionValue(sys, stateVector, inputVector)
f = sys.functionValue(x, u)

Description

functionValue evaluates the multilinear state equation of the model at the point $(x,u)$ :

$f = \Phi_F\, m(x,u),$

where $m$ is the norm-1 monomial vector over the signal vector $v=[x;\,u]$ and $\Phi_F$ is the state-equation block of the parameter matrix. What the result means follows the model’s timeStepSize: it is the state derivative $\dot x$ for a continuous-time model and the next state $x_{k+1}$ for a discrete-time one. Only the state equation is evaluated — use outputValue for $y$ .

Only the monomial columns the state block actually uses (columnIndexStateEq) and the rows stateEquationIndex are used, so the cost scales with the state block, not with the whole tensor.

Two evaluation paths exist:

CPNTensor — the unified value primitive computeFunctionValue, which compiles the sliced tensor and evaluates it in the MTISIM kernel.
Any other mtiTensor, e.g. TTTensor — the tensor’s own computeFunctionValueMss.

On the CPN path the signal vector is assembled by scattering $x$ and $u$ through stateIndex and inputIndex — exactly as stateJacobian and outputJacobian do — so a permuted structure-matrix row layout is evaluated correctly. For the natural [x; u] layout that rmss, ss2mss and the connection routines produce, the scatter reproduces [stateVector; inputVector].

Input arguments

Argument	Description
`sys`	The `mss` model.
`stateVector`	State $x$ at the evaluation point (`sys.nState` × 1).
`inputVector`	Input $u$ at the evaluation point (`sys.nInput` × 1; `[]` if the model has no inputs).

Output arguments

Output	Description
`functionValues`	`sys.nState` × 1 vector: $\dot x$ for a continuous model, $x_{k+1}$ for a discrete one.

Example