Autonomous mobility relies on cutting edge engineering and artificial intelligence to drive safely. Less than twenty years ago, many self-driving vehicles could travel no faster than a walking pace. Fast forward to today, and you’ll find self-driving developers like Uber ATG working hand in hand with organizations like Edge Case Research to co-develop safety practices that enable test platforms to comfortably navigate streets without human intervention. However, even with this progress, a great deal of work still lies ahead for the industry.
In 2019, Uber ATG published and open-sourced a first-of-its kind Safety Case Framework. A safety case, tailored from the safety case framework to achieve the goals of a program or product, includes a structured set of goals, argument, and evidence supporting the proposition that a self-driving car is acceptably safe for use in the real world. The use of a safety case does not mandate the use of any specific technology in creating the self-driving system. Safety cases are used in safety-critical applications such as aerospace, rail, nuclear and even healthcare. But they also provide significant flexibility, which is critical for emerging autonomy technology. The flexibility of safety case argumentation allows work products from traditional functional safety standards to stand alongside claims about the safety of novel approaches such as deep learning.
At the time of Uber’s Safety Case Framework debut, Uber ATG’s Eric Meyhofer stated, “We also recognize new or refined best practices may emerge over time, and we will consider these for incorporation into our Self Driving Vehicle (SDV) Safety Case Framework as they surface.” In the months since its release, Edge Case Research conducted a detailed review of Uber ATG’s safety case framework against UL4600 with the goal of producing a new, refined version of their framework. To do this we draw on Uber ATG’s experience maturing its own SDVs, as well as Edge Case Research’s experience providing safety engineering solutions across the autonomous vehicle industry, and our shared experience developing ANSI/UL 4600 and other voluntary industry standards.
New standards such as ANSI/UL 4600, Standard for Safety for the Evaluation of Autonomous Products, provide guidelines for how self-driving vehicle manufacturers can evaluate their own safety cases in a consistent way. Beyond the static assessment of a safety case at a single point in time, these guidelines are intended to establish the continuous risk analysis of self-driving vehicle fleets. By monitoring industry-accepted safety performance indicators, we can also accomplish this in a way that protects developers’ sensitive and proprietary data. Continuous risk analysis will help provide the public with confidence in self-driving car technology and support smarter insurance underwriting in the future.
In addition to peer reviewing Uber ATG’s Safety Case Framework, Edge Case is also working to productize tooling necessary to construct and maintain safety cases for large-scale operations. Edge Case’s vision is to provide safety case frameworks to customers along with the tools to validate these safety cases in the real world. These tools include Hologram, which SDV developers use to improve the safety of their deep learning perception algorithms, as well as new technologies in development at Edge Case. Edge Case is also working on safety case frameworks for developers building other kinds of autonomous vehicles, such as trucks, package delivery robots, and shuttles.
While Uber ATG’s Safety Case Framework was written to address fully driverless operations, the ATG team is wholly committed to tailoring the framework to achieve interim milestones and projects that lie ahead on the journey to self-driving vehicles. By aligning workflows to satisfy the safety case at every stage of development, Uber ATG is well positioned to utilize available resources and gauge progress towards safe self-driving ridesharing. Since a fully developed safety case will require several thousand pieces of evidence, Uber ATG continues to invest in tooling to support the curation and tracking of safety case evidence. With robust tooling, Uber ATG will be able to track and achieve multiple safety cases for multiple internal projects and milestones in tandem.
In the future, autonomous mobility will transport us to work, carry our families and friends to see each other, and deliver the goods we count on every day. The goal of Edge Case Research and Uber ATG’s collaboration is to ensure that everyone stepping into a self-driving car gets a safe ride, and that every self-driving vehicle traveling through our neighborhoods is built safely from the ground up.