I’d like the next autonomy team to spend less time reconstructing basic safety analysis, and more time finding what was missed.
That’s why I helped start the Open Source Safety Consortium (opensourcesafe.com), with Polymath Robotics as a founding member.
We’re developing reusable evidence for open-source components: hazard analyses, safety requirements, tests, assumptions, and known gaps, versioned alongside the code. Our first worked example is an open-source protective-stop prototype, with firmware, hardware designs, and a draft safety case all getting published together. We will then expand into other work asap.
It is not certified (yet!) and requires careful testing before using it in production (Polymath does use it). We are targeting the rigor required of a SIL 3/PLr=e capable solution.
Reusing its evidence would not automatically approve a machine, for example a haul truck still needs its own safety story around braking, and mine-site validation.
The question we want to test is how much another developer can actually reuse, what still needs independent review, and whether the effort saved outweighs the work of adapting and maintaining the evidence.
If you’ve qualified autonomy for a use case, what did you have to recreate because earlier evidence wasn’t available? And what would a shared package need before you or your assessor could use it?
If this all seems interesting to you, check out the above website, and send a hello!