Preparing a SALT Campaign

The system specification defines the data required for the SALT workflow. It includes the input variables, output variables, constraints, model configuration, and safe center point of the system under test.

Input variables

The input data must include the input variables of the system under test. The input variables describe the quantities that can be varied during the measurement campaign.

For each input variable, the permissible range and the permissible gradient must be defined. These definitions ensure that generated trajectories remain within the physically meaningful and executable operating range of the test bench.

Output variables

The output variables define the measured system responses. For each output variable, a NARX model is configured to represent the temporal system behavior.

Internally, the models use an ASCMO Gaussian Process with a Sparse Constant Sigma formulation. The models provide predictions and uncertainty estimates for the iterative online learning process.

Safe center point

A safe center point must be defined. The safe center point is used as the initial operating point and as a fallback position during the exploration of the input space.

The safe center point should represent a known safe operating condition for the configured inputs and outputs.

Constraints

The system specification can contain input constraints, an input validity constraint, and output constraints.

Input constraints define the permissible operating region based on existing domain knowledge. These constraints can be imported from an ExpeDes project, for example to represent speed-load limits.

The input validity constraint bounds the overall exploration space to the physically meaningful input domain.

Output constraints monitor safety-critical output variables. They are used to avoid predicted limit violations during the generation and execution of trajectories.

The input data, constraints, and model configuration are stored in a SALT configuration project file.