eventTotalDerivativeTT

Method of continuousTimeMdss.

Consistent initial derivatives for a tensor-train (TTTensor) mdss model.

Source: src/simulator/@continuousTimeMdss/eventTotalDerivativeTT.m

What is computed

eventTotalDerivativeTT is the tensor-train counterpart of eventTotalDerivative’s initial solve, for the non-hybrid path (no binaries). Given a state $x_0$ and the current input value and slope $(u_0, \dot u_0)$ , it produces a consistent initial point for ode15i.

First it solves the implicit equality system for the state derivative $\dot x_0$ and the algebraic signals $y_0$ ,

$F(\dot x_0, x_0, u_0, y_0) = 0,$

with lsqnonlin (Levenberg–Marquardt) using the analytic Jacobian. It then back-solves a consistent algebraic slope $\dot y_0$ from the equality Jacobian for the fully-implicit solver; if that back-solve is not finite, $\dot y_0$ is returned as zeros.

An empty yGuess defaults to zeros.

Input arguments

Argument	Description
`sys`	The `continuousTimeMdss` simulation object (its `mtiModel` is TT-backed).
`x0`	Current state values.
`yGuess`	Initial guess for the algebraic variables; `[]` defaults to zeros.
`u0`	Input values at this time.
`du0`	Input time-derivatives at this time.

Output arguments

Output	Description
`xp0`	Consistent state derivatives $\dot x_0$ .
`y0`	Solved algebraic variables.
`dy0`	Consistent algebraic slopes $\dot y_0$ (zeros if the back-solve is non-finite).

eventTotalDerivativeTT

What is computed

Input arguments

Output arguments

See also