A repository for models and source code for probabilistic safety estimation via monotonic safety (MoS) and MoS-based abstractions.
The directory structure is as follows:
- code: This directory contains the code for generating and running the models, for both perception and control, along with code for generting counter examples to the monotonicity of safety rule.
- models: This directory contains the generated models accompanying the paper. It is separated by perception models and control models, for both case studies.
- results: This directory contains the data and visuals accompanying the paper, along with the code used to generate the the visuals.
- old: old and miscellaneous files. This directory can be ignored.
All of the perception bundling, some visualizations, and some experiments are implemented on the Wolfram language/Mathematica stack. It has different parts described below.
Requirements: Wolfram Script version 1.5+.
To run the scripts in this software package, you need to dowload and install a free version of the Wolfram Engine:
- Download and install the Wolfram Engine for your OS.
- Create a Wolfram ID account (if you do not have one).
- Accept the terms of the free license
- Run "wolframscript" in your command line
- Input your Wolfram ID and password into the prompt
The scripts are files with .wls extensions that generate bundled perception distributions. Their output is typically sent to the command line or in a set of files, depending on the script.
All scripts are standalone and run as follows:
wolframscript <scriptname.wls> <arg1> <arg2> ...
The notebooks implement various exploratory experiments, visualizations, and cached bundlings.
Requirements: For editing access, you require Wolfram Mathematica, version 12.1+. You can obtain a free 30-day trial or you find yourself be at an instutition with an academic license. For read-only access, you can download a free Wolfram Player.
Instructions:
- Open a desired .nb notebook,
- Either run it all at once (Ctrl+A, Shift+Enter) or line-by-line (select a line, Shift+Enter).
- Follow the instructions on https://carla.readthedocs.io/en/0.9.11/