TL;DR: On a large cruise ship the GPS antenna broke, and the GPS continued to display plausible coordinates based on dead reckoning with only a tiny icon to indicate the data wasn't based on a satellite fix. The dead reckoning error accumulated to 15 miles, which drove them aground near Nantucket.
Actually I think the real TL;DR is that the crew failed to "trust but verify" and that the automated cross checks were set up in such a way that they cancelled each other out. It's a cautionary tale for anyone developing self-navigating devices.
It's a copy of a book chapter, plus the end notes (notice how the page number jumps near the end). The end notes aren't paginated by chapter, so you're seeing the first paragraph of the end notes for the NEXT chapter at the bottom of the page.
More than that, the dead reckoning data from the GPS "infected" numerous other automated systems that should have caught the error, because there was (is?) no defined standard for what a GPS receiver should do in the event of losing its fix. In this case, it kept transmitting position data and added a proprietary "invalid" tag that wasn't parsed by other equipment.
:-) One of my uncles has just retired from the main UK collage for training Merchant Marine officers - he commented that a lot of the younger officers don't have much experience of using the traditional methods of fixing your position.
This is really well written, the author has a very engaging story-telling style, and manages to get the facts across while still being entertaining. And, I love the Proverbs quote at the start. Anyone know what book it comes from, as I'd love to read it if all the content is like this.
It appears to be the entirety of Chapter 8 of something, but I couldn't find the actual book title mentioned anywhere...
Degani, A. (2004). Taming HAL: Designing interfaces beyond 2001. New York: St. Martin’s Press/Palgrave Macmillan. 320 pages, 105 figures. ISBN: 031229574X
The book details a new approach, mostly based on software engineering and formal methods, for the describing, analyzing, and identifying potential design errors in human-automation interfaces.
