Delivering Live Safety Cases for an Ever-Changing World
Michael Wagner, CEO, Edge Case Research
Self-driving cars will need to operate in an ever-changing world without the benefit of flexible, adaptive human drivers continuously overseeing their operation. No design team can foresee and plan for every potential safety risk — much less the changes that will occur in the environment over the months and years that a self-driving car operates — without taking a fresh approach to safety.
That’s where Edge Case Research comes in.
We enable our customers to assure the safety of autonomous systems for real world deployment in the face of complex, changing operational environments. We do that by developing products that deliver live safety cases backed by evidence to ensure that our customers can define and measure the safety of autonomous systems in the real world.
A safety case is an organized way to ensure that engineering, testing, and safety culture all contribute to a safe vehicle. All claims outlined in a safety case are backed up by evidence. However, Edge Case’s approach to safety cases goes a step further with live safety cases.
Traditionally, safety cases are often fixed documents with static links to reports. In contrast, a live safety case processes data from simulation testing, on-road testing, and field operations to support continuous improvement and re-validation. This live aspect of the safety case is essential to address safety risks and other surprises that are sure to appear during operations — and to address them before a crash has a chance to happen.
Live safety cases are built through collaboration among safety engineers, systems engineers, developers, and other stakeholders. Next, sources of live data are connected to the structure of the safety case itself. This establishes a positive feedback loop that scales risk analysis, improves safety requirements, and conforms to standards such as the Safety of the Intended Function (SOTIF). Over time, this positive feedback loop gets the vehicle to a point at which it is safe enough to deploy.
But that’s not the end of the safety story.
When self-driving cars are put on the road, they continually encounter “edge cases,” unusual situations not considered in the design and not seen in pre-deployment testing, that present unforeseen safety risks. They might be caused by changes in everyday traffic flow such as road construction or crossing guards appearing at the start of the school year. They might occur when traffic laws or signage changes. Effects from unprecedented weather and changes in fashion might also pose problems. Live safety cases give you a chance to identify these problems in data from field operations and resolve them before they lead to a crash.
Our goal is to develop products that deliver live safety cases to our wide range of customers, such as OEMs building advanced vehicle safety features, networks dispatching self-driving cars and trucks, operators of mines and farms, and aerial drone delivery service providers. Our live safety cases are built to conform to standards and best practices, and to deliver the expertise needed to build and operate safer and reliable autonomous systems globally.
To learn more about our product development, supporting live safety cases, please email us at info@ecr.ai
