psy-taliro 1.0.0

Search-based test generation framework

$\Psi$-TaLiRo

An extensible Python toolbox for search-based test generation for cyber-physical systems. This work is based of the MATLAB toolbox S-TaLiRo, which is available here.

Models

$\Psi$-TaLiRo represents the cyber-physical system under test as a model. Conceptually, a model is a function that maps a set of system inputs to a timed series of states called a Trace. $\Psi$-TaLiRo provides two ways to construct a model: the Blackbox and Ode models. The Blackbox model makes no assumptions about the underlying system it represents. The Ode model expects the underlying model to be represented using ordinary differential equations.

Specifications

$\Psi$-TaLiRo tests are system requirements expressed in metric temporal logic (MTL). Evaluation of the system requirement depends on a monitor, and $\Psi$-TaLiRo supports several options. All specific implementations are available in the staliro.specifications module. To use a specification, you will need to ensure that the monitor library is installed. Additional information is available in the table below:

Monitor	Installation	Link
RTAMT	pip install rtamt	homepage
TLTK (*Linux only)	pip install tltk_mtl	homepage
Py-TaLiRo	pip install pytaliro	homepage

Optimizers

$\Psi$-TaLiRo generates inputs for the system under test by using an optimizer. The optimizers provided by the toolbox are the UniformRandom and DualAnnealing optimizers in the staliro.optimizers module.

For other optimizer options, see the PartX repository. There you will find additional implementations that give extra guarantees about the system input space.

Type hints

This toolbox provides PEP484 type hints to help ensure correct usage. To use the type hints, you will need to install one of the several type hint checkers available for python. A non-exhaustive list is:

• MyPy
• Pyright
• Pyre
• Pytype

The easiest way to get started is to install VSCode with the python extension which includes the pyright type checker.

Installation

To install this toolbox, run the command pip install psy_taliro. To avoid installing python packages globally, you can use a virtual environment to keep the packages in a project-specific directory. Some of the tools for managing virtual environments are:

• Pipenv
• Poetry
• Venv

(Anaconda)[https://www.anaconda.com] can also be used to create separate python environments, and may be easier to set up on some systems.

Example

from math import pi

from staliro import models, optimizers, specifications
from staliro.options import Options
from staliro.staliro import staliro


@models.blackbox()
def aircraft_model(X, T, U):
    """Blackbox model that represents the dynamics of an aircraft.

    Arguments:
        X: The static (initial) inputs to the system. A four-element vector of
           the form [roll, pitch, yaw, thrust].
        T: Interpolation times for the input signals
        U: Interpolated values of the time-varying input signals

    Returns:
        trace: A set of timed state values representing the altitude of the
               aircraft over time
    """
    ...


optimizer = optimizers.UniformRandom()
requirement = "[] (alt > 0.0)"  # Requirement that the aircraft does not crash
specification = specifications.RTAMTDense(requirements, {"alt": 0})  # The altitude value is in the first column of the aircraft trace states
options = Options(
    runs=10,  # 10 independent optimization attempts
    iterations=100,  # Generate 100 samples per optimization attempt
    interval=(0.0, 2.0),  # Simulation interval is from 0 to 2 seconds
    static_parameters=[
        (-pi / 4, pi / 4),  # Roll
        (-pi / 4, pi / 4),  # Pitch
        (-pi / 4, pi / 4),  # Yaw
        (0, 100),           # Thrust
    ]
)

result = staliro(aircraft_model, specification, optimizer, options)

Documentation

For additional details about the toolbox components, or example scripts, refer to the documentation site.

Citing this project

If you use this toolbox in your research, include this citation in your bibliography

@misc{psy-taliro,
  doi = {10.48550/ARXIV.2106.02200},
  url = {https://arxiv.org/abs/2106.02200},
  author = {Thibeault, Quinn and Anderson, Jacob and Chandratre, Aniruddh and Pedrielli, Giulia and Fainekos, Georgios},
  keywords = {Software Engineering (cs.SE), Systems and Control (eess.SY), FOS: Computer and information sciences, FOS: Computer and information sciences, FOS: Electrical engineering, electronic engineering, information engineering, FOS: Electrical engineering, electronic engineering, information engineering},
  title = {PSY-TaLiRo: A Python Toolbox for Search-Based Test Generation for Cyber-Physical Systems},
  publisher = {arXiv},
  year = {2021},
  copyright = {arXiv.org perpetual, non-exclusive license}
}