5 ms·
Really good clear explanation of the issues. The group I work in (Amey Consulting) have built a couple of systems for Network Rail here in the UK, called Quart
by steve_gh 4y ago
Really good clear explanation of the issues.
The group I work in (Amey Consulting) have built a couple of systems for Network Rail here in the UK, called Quartz and Cosmo. These are analytic systems examine rail delays and delay propagation, especially sub-threshold delays (in the UK delays over a threshold - 2 mins IIRC) have to be reported, but those under 2 mins don't. We started looking at sub-threshold delays a few years ago, and have understood how these small delays build and propagate through the network, affecting multiple services.
We are always recruiting (UK - main sites London and Manchester, some folk in Birmingham too) - so if you have an interest in applying data science to infrastructure have a look here: https://www.amey.co.uk/what-we-do/amey-consulting/careers-with-amey-consulting/careers-in-advisory-analytics https://www.amey.co.uk/what-we-do/amey-consulting/careers-wi..., or DM me
- zelos 4y agoOut of interest, what kind of state is software engineering in the rail industry in? Watching a documentary about Crossrail the other day their testing seemed to consist of a driver manually driving a simulator in real-time through signals and recording when buttons don't work etc. I was kind of hoping that was just a misrepresentation.
- Milner08 4y agoI assume part of the reason for doing it like that is that it gives user feedback for the systems rather than just following what ever assumptions they put in place for their automated testing. It also servers to be driver training, which is very important and takes a long time, so if they can start sooner its probably a good thing.
- lbriner 4y agoI saw the same program and was cringing. "The train has crashed" now meaning its software has crashed and it needs a reboot. It is not frightening that it might physically crash the train, which I assume has crazy numbers of failsafes but more about the terrible reliability of some software which has unexpectedly crashed despite whatever modelling/testing was already carried out. If it is that unreliable, it has to have loads of failsafes which means lots of unexpected failures. At least with older trains, you could usually move the train in emergency, even if slowly, but having the software tell you that you can't close the doors or pull away is really bad. There were also issues with trains missing from the passenger information systems and even worse, a broken down train on the boundary between TfL-controlled and Network Rail-controlled lines and it sounded like no-one knew what to do and it caused 2 hours of messing around.
- Mvandenbergh 4y agoInteresting. Do you find that the mathematics of delays shows soliton properties? I know that highway traffic does.