Predictor

Class for implementing model-based prediction using sampling.

This class defines an obect for performing model-based state prediction using sampling methods. A Predictor is constructed using a PrognosticsModel object, i.e., it must have a state equation, input equation, and threshold equation defined. The states are simulated until either a specified time horizon is met, or the threshold is reached, as defined by the threshold equation. The input equation is used to compute the inputs to the system at any given time point. The input equation should be defined to take in addition to the time, a set of "input parameters" that are used to determine what the input should be at a given time.

The predict method takes the starting time of prediction, the horizon to predict to (i.e., it simulates to starting time + horizon), and functions to generate samples for the state at the time of prediction, the input parameters, and process noise. If horizon is inf, then prediction will continue until all samples hit the threshold. This is not recommended unless it is guaranteed that each sample will reach the threshold.

The sample generator functions must be defined to each take only one argument, the number of samples to generate. This means that whatever information is needed to correctly generate samples should be built into these functions, such as distribution parameters.

Properties

• predictions - Data structure containing results of the predict method.

Methods

• predict - Predict the state evolution up to a threshold.

See also Prognoser.