Workshop

Consider the following scenario for a highly automated system:

We are in court.  Our product has failed.  (The truck has crashed, the plane has exploded, the robot has crushed someone.)  We did everything possible to try to prevent this from happening.  How do we prove our case and show that we did what we could?  What evidence and artefacts do we present?  How do we evaluate them?

In court we would have to show that we adopted best practices and principles such as ALARP for risk reduction (as low as reasonably practical). This, and the above questions, are particularly challenging now that automated systems are breaking new ground, with new functionalities/sensors/computing being applied in complex open environments, (typified with automated vehicles on the street, and as opposed to for example, the caged robot).

This means that there are limited best practices available … existing safety standards still assume the presence of humans and provide very little guidance when it comes to how to ensure safety of complex functionality for perception (environment understanding), AI and deep learning.

In the workshop, perspectives to safety assurance will be provided and we will be organizing a world-café (see below) to discuss challenges and promising approaches to deal with them.

Preliminary program:

Inspiration talks:

• Invited talk on Safety assurance for highly automated systems – John McDermid, York  (25 min)
• Providing tool support for safety assurance: perspectives
  from the automotive industry, Sofia Cassel, Scania (15 min)
• Research challenges on safety assurance – Martin Törngren, KTH (15 min)

Leg stretch

World cafe sessions with the following themes/tables and a moderator/table

• Safety cases for safety assurance of highly automated vehicles (Damir N., KTH)
• The role of scenarios/assumptions and architecting for safety assurance (Naveen M./Xinhai Z, KTH)
• Dealing with risk and uncertainty (Fredrik A./Masoumeh P., KTH)
• Other topics including Forensics – how do we know what the problem was and how the system reacted (Martin T, KTH)
  
  ROTATIONS among the tables. E.g. 20 mins per table should allow a full rotation

Summaries from tables and wrap up

The approach of the workshop is to provide a rich forum for discussions. Short presentations will provide summaries of positions and ideas. Subsequent discussions will be structured as world café sessions. As a world café, discussions will be divided into thematic tables, each with a chair person. Participants will rotate among the tables. The role of the chair is to structure the discussions, to summarize findings in a follow-up plenary, and to write a summary,

http://www.theworldcafe.com/key-concepts-resources/world-cafe-method/

Instructions for Theme/Table chairs

Each theme has a corresponding table and a chair. The role of the chair person is to (1) structure the discussion at the table and (2) to provide a short summary of the discussions (of all rounds for the table), to be provided at the plenary.

To lead the discussion, the table chair is asked to prepare a couple of questions. It is useful to prepare a couple of questions beforehand – and then to extend these given the initial presentations and material made available before the workshop. When starting the theme you can ask if there are further suggestions for such questions. The discussion would then be structured using the questions.

The chair is expected to briefly summarize the results from the table session discussion at the plenary. This is preferably done in some written form (e.g. pptx) to be able to easily collect these summaries later on.