Simulation Accelerator Toolbox for MATLAB

This toolbox presents an abstraction layer that can be put in front of our Subset Simulation Toolbox to simplify and speed up subset simulations with Simulink models. In short, it contains a custom code generation and compilation process that produces a binary simulation target that is optimised for speed in the context of subset simulation.

The details of the inner workings and architecture are captured in the user manual and publication linked above. In short summary, it might give you 5–15⨉ faster simulation results when you iterate Monte Carlo samples than what you are able to achieve using the standard simulation tools.

Features

Simulation target optimised for speed

Contains our CompiledScenario class that produces a binary simulation target that is optimised for speed when used together with our Subset Simulation toolbox.

Easy setup of subset simulation studies

Define a simulation Scenario, a list of Parameter objects and a failure Metric, then join them in a subset simulation Study that is ready to start.

Iterative build artifact generation

Our custom code generation procedure uses an iterative Makefile-like approach. To regenerate certain artifacts, just delete them and start the process again.

Support for many official toolbox features

Use embedded or external test harnesses, load scenario configuration from Test Manager files, automatically discretise models, and automatically select the best call signature of sim() and parsim(). If you are familiar with the official APIs, you should be able to get started quickly.

Case Studies

Speeding up simulation of an eVTOL

In this study (DOI link below), the authors compared simulation times for an eVTOL simulation model of comparative modelling complexity as would be needed for requirement verification. The code generation toolchain available in this toolbox (in the table named “proposed”) had a comparable setup time as the other simulation modes, but execution time was only a fraction, leading to a simulation speedup of 73⨉ in total, or 5⨉ above the fastest built-in simulation mode (rapid accelerator).

doi:10.2514/6.2021-1980

Contact

If you have questions, feedback, bug reports or issues regarding licensing, please contact us at: