Hacker News new | past | comments | ask | show | jobs | submit login

So, a modern aircraft probably have a few thousands sensors (if not way, way more), not strictly redundant but overlapping in some fashion -- I have to believe there's been some modelling trying to check that sensor input is consistent; if only to have some kind of way for the computer to figure out that a sensor has gone crazy (say your compass suddenly detects a significant change in heading, but no sensor has detected even a degree of roll, a change in speed, heading computed by GPS over time is stable... compass might be borked, whaddya say?)

Why isn't such modelling used to let sensors be disabled when they're clearly sending bogus input?




My guess is that such systems are complex, hard to model for safety assurances, and hard to reason about in the moment.

Yes, it will probably make the system safer, but your guarantees become a lot more fuzzy. This is not the field for fuzzy guarantees.


> Why isn't such modelling used to let sensors be disabled when they're clearly sending bogus input?

You need to crash a lot of planes to train that NN.


there should be at least 3 AOA sensors - so you can not only detect a mismatch, but figure out which sensor is most probably incorrect.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: