ai4safety Roadmap

Intention of this page is to drive awareness on our intentions and our direction in this project.

Why are we Here?

I think it’s good to get this right out into the open and transparent with each other so we know where each us are coming from.

Name

Altruistic Reason 😇

Selfish Reason 😈

Name

Altruistic Reason 😇

Selfish Reason 😈

@Olivier Cornes

I want to develop the field of computational safety engineering so that complex safety critical systems can be built more easily.

I want to get a PhD out of this research work that will allow me to teach at university level in Switzerland and Central Europe.

@Michael Mustillo

I never want to see something as preventable as the 737MAX, ever again!

Practical experience in application of AI to boost my portfolio in industrial applications.

@Esma Karagöz

I would like to advance the research in the field of AI and MBSE.

I need to improve my publication record.

@robert karban

 

 

AI4Safety Objective

Boiler plate, to be refined:

Our mission is to increase the robustness of safety analysis in vehicle systems, with the goal of making them easier to assess, and thus easier to ensure their safety in operation.

By integrating Large Language Models (LLMs), we aim to identify and bridge safety gaps, preventing oversights akin to those in the 737 MAX incidents. Our interdisciplinary team is dedicated to advancing safety standards through innovative AI applications, promoting reliability and integrity in critical system operations

Roadmap

This section provides a visual and tabular roadmap.

The visual roadmap will dictate the direction we want to take at a high-level, identifying key interim yet publishable accomplishments on the way to our final achievement.

What are our interim goals? What is the final goal, given our eventual objective?

image-20240227-205715.png

The tabular roadmap will decompose this into further details, providing a starting strategy for each goal.

#

Goal

Plan of Attack

Target Conference

Abstract Submission Deadline

Paper Submission Deadline

Conference Event Date

#

Goal

Plan of Attack

Target Conference

Abstract Submission Deadline

Paper Submission Deadline

Conference Event Date

1

Use LLMs to demonstrate repeatably accurate knowledge graphs of aircraft systems.

  • Gather examples of aircraft data for LLM analysis

  • Create their ground truths

  • Analyze the examples with 3 major LLM types (ORCA2, GoLLIE, and Mistral)

  • Validate reliability and accuracy of outputs

  • Create knowledge graphs

MODELS 2024 - Technical Track - MODELS 2024 (researchr.org)

Mar 21, 2024

Mar 28, 2024

Sep 22, 2024

2

 

 

 

 

 

 

3

 

 

 

 

 

 

4

 

 

 

 

 

 

5

 

 

 

 

 

 

Project Management Ideas

  • Use Jira for tracking progress

  • To keep it simple, one sprint per goal, identified as epics

  • Weekly meetings on Tuesdays at 10am PST, 11am CST, 6pm CET

  • Identify Clear Roles and ensure tasks are well defined and understood

  • No stomping on each other’s tasks

  • Fairness with authorship