You can think of it as a flowchart that splits into hundreds of boxes, and a self-driving vehicle must pass each to be declared safe. Aurora explains that its tool is industry-first because it is both for trucks and passenger cars.
Regardless, the first version that was released is just a first step in this framework, as Aurora plans to expand it to new environments, scenarios, and platforms.
Aurora is already working on a system that enables self-driving capabilities for passenger cars and trucks. They called it the Aurora Driver, and it will be based on this framework.
Aurora claims the top four levels of the framework are already shared across the industry because the company is confident in its progress. The five pillars of the framework are as follows: Proficient, Fail-Safe, Continuously Improving, Resilient, and Trustworthy.
The first pillar is focused on assessing if the self-driving vehicle is "acceptably safe" during nominal operation. The second pillar gauges if the same level of safety is achieved if faults and failures occur.
Meanwhile, the third pillar of the framework is focused on eliminating previous errors, if risks are evaluated, and how those issues are resolved to prevent them from being repeated.
Moving on to the fourth pillar, which targets foreseeable misuse and unavoidable events. The target is also an acceptable level of safety in all those situations. The last pillar's purpose is to determine if the enterprise is trustworthy.
Time will tell if this framework is enough or just the first step in developing an assessment method to gauge the safety of autonomous vehicles. As Aurora points out, trustworthiness is the last step, and what the framework cannot figure out is what it would take for humans to trust self-driving vehicles.