On Jan. 24, another F-35 crashed when landing on the aircraft carrier Carl Vinson after another crashed while taking off from aircraft carrier HMS Queen Elizabeth & a few more. These misfortune incidents may not because of a single root cause. And, no one is an easy problem to solve. Nevertheless, we should have the decency to push the boundaries to tackle them better.
Point of no return
Software crisis has been a thing since 1968, despite the technology & business have been improved multiple folds since that. However, there are still way too many SW projects taking much longer & costing too much. F-35, as the most advance fighter jet is also a subject of such crisis, and it’s not an isolated case because the complexity grows faster. By the rate, maybe there is no escape from the complexity black hole, right?
SW seems to be the black hole sucking all additional resource & time. Lockheed Martin delivered 142 F-35 in 2021 & aims for 151 in 2022 as Reuters reported. However, 1/3 of them may hit the sky before fully tested according to US Government Accountability Office. The delay is mainly because of holdups in developing an aircraft simulator. Now, there is increasing worry up to the Congress that the modernization will not complete in 2027 & under $2T even after a few delays & a few more billions added.
To be fair, F-35 Joint Program Office(JPO) has been trying very hard to recover, such as adopting Agile software development & Continuous Development to Delivery to release increment changes every 6-months since 2018. Just for a bit of comparison, top mobile apps typically update every 1.5-months, and Android phones update about 3-months. I bet F-35 modernization program has got most points on testing early at scale and even reduce bugs by augmented reasoning. These are the good ways to get confident on the code changes. So what else one may do better?
Metaverse of salvation
Even if all code changes are good, there can still be problems on the complex variants of data, situational & state in the real world. Therefore, JPO pursues a good strategy: “simulation difficulties”. Because, live training events can not cover all possible complex & highly advanced threats. Furthermore, test flights costs too much, takes too long & adds more risk at scale & a high frequency. Simulation may be the only sustainable way to go.
Nowadays, not only modern fighter jets have to simulate. Tesla also hopes to simulate a way out to reach 99.9999…% of self-driving in real life. Sure, simulation is just a proxy. But, only relying on user trails is the worn answer even if anyone can afford it. For example, mobile phone makers & users often frustrate because developers cannot reproducible the bugs to fix them. So even building & running high reliable simulations is not easy, it does scale better. The question is: how may the one be built quicker & better?
Taming madness of multiverse
Obvious the “simulation difficulties” strategy is not easy to deliver because of “late contract awards preventing them from conducting new work, supply chain issues…” according Lockheed. So, how may one do better? What if?
- Each SW components & high fidelity HW emulations can interact with each other to validate corner cases end-to-end. With no HW dependency, vendors can start early with little initial cost. Also, each vendor can validate their changes against known goods & ensure their regressions will not propagate to downstream.
- The meta data set & situation records are aggregated & shared across makers. As the meta data as “the interaction contracts” defined, everyone can start their own development iterations in parallel instead as a traditional serial development pipeline. Furthermore, all share the lessons learned & easy to gate their changes from repeating the mistakes.
- Offending cases can be easily replied for anyone to reproduce & debug. So, less time on asking for more info, reproducing & pointing fingers.
Full Disclosure
The opinions stated here are my own, not those of my company. They are mostly extrapolations from public information. I don’t have insider knowledge of those companies, nor a whatever expert.
