RailPilot - Journey Tracker

[chr15toffer]

Hi all

My name's Chris and I've been working on a concept for a web app called RailPilot that fills some of the void that I see between sites like RTT, OTT and the journey planners. Fed up with having to juggle multiple tabs to see whether I'm actually going to make a connection, I've devised something that to me is genuinely useful, but it would be great to get some feedback from people that really understand how the railways work.

Some of the key features:

• Tracking connections in realtime, highlighting alternatives where connections are tight and replanning automatically when they're missed
• GPS matching to services and stations so if the user is already travelling they only need to declare their final destination and the app works out the rest
• In-journey narration, including delay reasons and status of signals when services are held
• Machine learnt punctuality, formation and delay data, giving the user the ability to make a more informed choice about which journey to take (e.g. avoiding services that are frequently cancelled or short-formed)
• Presenting status of the forming service, allowing users to replan before Darwin admits there's a problem
• Realtime assessment of services ahead of the user's, informing them if they're contributing to a delay (e.g. calling out a late running stopper) and exactly when that service will clear the user's route
• Learned delay data overlaid on to Darwin and TRUST information, flagging typical delays given the service's current position and scenario
• Ability to share journey progress in realtime, for example with someone coming to collect you at your destination

I've done a good amount of testing, but this is very much still in beta so you may encounter issues and edge cases.

If you think it's awful then please be honest, if you think it's great then please say and if you have constructive feedback as to how it could be better then it will all be very welcome.

Thanks,
Chris

 

[Vexed]

Presenting status of the forming service, allowing users to replan before Darwin admits there's a problem

Darwin should already do this, though it relies on operators setting things up properly.

On the face of it, it looks impressive and could be pretty helpful. I hope the backend is well written and tested as the Darwin feeds have their quirks and edge cases, and not just blindly trusted AI output. I've written a simpler thing for watching connections myself, but I've kept it private as it has a few issues that need quite a lot of development time before I'm happy releasing it (and I was waiting a while for RDM to fix some problems on their side).

I've just tried setting my location in DevTools (not your UI) to Worthing but it seems to snap to assuming I'm on a train between Southwick and Portslade?

The rail industry could help a lot here with GPS data. I'd expect quite a few "normals" get what train they're on mixed up on so more open data would be really helpful. My understanding is that most trains do report live coordinates to an industry system called Customer Informaton GPS Portal (CIGP) but this data isn't available publicly. Therefore I'd assume there's a lot of guesswork in your code. Even the TD feed (the most granular location data publically available) doesn't have any official mapping to coordinates so I think you're sort of trying to do the impossible here, like signalbox.io that has its own issues.

I like how it shows a tight connection below MCT (Minimum Connection Time) as potentially possible.

Just some other thoughts, does this look at the Routeing Guide at all?

 

[chr15toffer]

Darwin should already do this, though it relies on operators setting things up properly.

On the face of it, it looks impressive and could be pretty helpful. I hope the backend is well written and tested as the Darwin feeds have their quirks and edge cases, and not just blindly trusted AI output. I've written a simpler thing for watching connections myself, but I've kept it private as it has a few issues that need quite a lot of development time before I'm happy releasing it (and I was waiting a while for RDM to fix some problems on their side).

I've just tried setting my location in DevTools (not your UI) to Worthing but it seems to snap to assuming I'm on a train between Southwick and Portslade?

The rail industry could help a lot here with GPS data. I'd expect quite a few "normals" get what train they're on mixed up on so more open data would be really helpful. My understanding is that most trains do report live coordinates to an industry system called Customer Informaton GPS Portal (CIGP) but this data isn't available publicly. Therefore I'd assume there's a lot of guesswork in your code. Even the TD feed (the most granular location data publically available) doesn't have any official mapping to coordinates so I think you're sort of trying to do the impossible here, like signalbox.io that has its own issues.

I like how it shows a tight connection below MCT (Minimum Connection Time) as potentially possible.

Just some other thoughts, does this look at the Routeing Guide at all?

Thanks for the feedback, I appreciate it. I'm sure there are more edge cases to be found, but I've spent a lot of time getting to know Darwin so can be reasonably confident that there aren't too many left hiding.

Daft question, but were you moving when you tried spoofing your location? The algorithm matches you to a service broadly as follows:
1. Where are you and are you moving (>5m/s)?
2. Which services are you near? Built from the CIF schedule with STP alterations and Darwin/TRUST actuals folded in.
3. From the list at 2, which train can get you to your destination?
4. From the list at 3, where were those trains in the last few seconds, based on Darwin and TRUST, forecasted forward to draw a line and estimate a possible corridor for time 'Now'.

You're right, it does have a degree of guesswork, but the principle is to fallback gracefully. If I can save a few clicks and be confident the user is on a specific train then great, if not then it's down to the user to make the final call, from a shortlist. I am working on some improvements that will increase confidence, including TIPLOC coordinates. SMART does map some of the TD C-class messages to timing points/stations and it would be possible to learn the others to build a very granular picture and significantly increase confidence.

No, this doesn't consider the Routeing Guide at all, it's very much 'how do I get there soonest/fastest?' without any ticketing constraints, as is the case with something like Google Maps.

 

[Vexed]

Daft question, but were you moving when you tried spoofing your location?

No, but it could just be the me changing the spoofed location manually confusing it.