The Sandbox: the AVSandbox knowledge hub

Autonomous Vehicle Safety Series (AVS #)

The industrial application of Automated Vehicles is becoming ever more feasible. As the excitement of innovation wanes, the safety and reliability become the focus. Delivering a reliable and safe AV service can be achieved, when a safety-oriented culture creates an environment where multiple design and analysis iterations can continuously improve the delivery of safety as a service. While part of the testing must eventually be hardware, the multi-physics modelling capabilities of Dymola, coupled with the realistic Digital Twin simulations of rFpro and Claytex’ developed libraries, combine to create an opportunity to deliver a complete Safety Estimation Framework.

This framework is based on multi-staged, automated Software-in-the-Loop architecture, whereby the supplier-provided Operational Design Domain [legacy work] and customer-provided Usage Heatmap serve as input, offering a reliability and safety estimate as output. It features multi-physics models of both the sensors and actuators [example], to deliver the most precise, virtual safety testing environment so far. However, on our way there, there are a number of challenges, that will be discussed in this AV Safety series.

State of affairs

It is not the first time an emerging technology poses challenges. Electricity, the automobile, and especially aviation and nuclear have a long history of delivering inherent safety, satisfying the public’s need for its service while keeping the public risk ALARA (As Low As Reasonably Achievable). Historically, after a period of chaotic innovation, the government steps in to introduce standards based on the dominant technology and lessons learned from the operational challenges.

This process is already taking place in the Autonomous Vehicle sector, as standardisation follows the forefront of innovation, facilitating the communication and collaboration between all parties. With regards to safety, however, there are decades of operational experience in the nuclear and aviation, that the automotive industry can learn from. By combining this with recent developments in game theory to describe and generalise traffic situations, we aim for a new level of AV Safety insights based on the rFpro and Dymola stack. Stay tuned for the AVS Blog series as we report our milestones.

Towards safe AVs

“To be feasible, an autonomous vehicle needs to be safer than an average driver” – it is a popular phrase but what does it mean.

An average motorist driving today has less than 0.05% chance of sustaining a traffic injury in a year. Assuming that one is a motorist for 50 years, it means that out of 400 people only one will ever be affected. However, the safest present automotive vehicle being tested boasts 1 disengagement per 30 thousand miles. There are no statistics, that would correlate likelihood of injury given a disengagement. Assuming this likelihood to be 2%, the AVs still need to be 100 times safer to be comparable to humans.

Figure 1: Re-Defining Automotive Safety | Image: Freestock-photos @Pixabay

Defining the basics

But can we make such assumptions? How to measure the risk of various AV faults? What mitigation systems are there? How is a motorway mile comparable to an urban mile? Furthermore, what is the definition of safety in the first place? How does it relate to risk and reliability? Most importantly, what are the current regulations and guidance frameworks, and how does it all affect the public?

Public health is at stake. There are weak assumptions that need to be addressed, to offer a reliable metric of safety. This is what we are working to address.

Written by: Dr Marcin Stryszowski – Lead Engineer

Please get in touch if you have any questions or have got a topic in mind that you would like us to write about. You can submit your questions / topics via: Tech Blog Questions / Topic Suggestion.