[RFD] Finding routes and services

[Steve]

I think some things can be made simplier, as your identifying them as single passengers. In 5 for example, those 5 options will be the same for every passenger, so surely they'd choose the same route. The only factor would be crowdiness. Such a thing can be made by simply spliting the passengers by percentage between what every stations are avaliable for their route.

"Try a few times to give another random destination, then "discard" the passenger."
That seems a bit odd to me. If there is no route to London, they won't spontaneously go to Paris instead. Again, it's much simplier to just attract the passengers to the station for routes which you have set up.

[PJayTycy]

"Try a few times to give another random destination, then "discard" the passenger."
That seems a bit odd to me. If there is no route to London, they won't spontaneously go to Paris instead. Again, it's much simplier to just attract the passengers to the station for routes which you have set up.

Did you read the link I posted to the wiki? Maybe, but I'll tell again why I vote for this option:

If we give them random destinations on the map, and immediately discard the ones that are unreachable, we will have a very tough early game (almost no passengers).

If we distribute them among the reachable destinations only, whenever we add new connections, the traffic will drop on our existing lines. If only a connection between city A and city B exists, 100% would take that route. When we add a connection between city A and C, the traffic would be split so the A-B connection would have only 50% of the traffic anymore. So, an increase of your network would not translate into an increase of passenger volume.

This 4c option is something in between. If we chose 4a or 4b, we have to account for the number of connections when calculating the number of passengers a city generates. This 4c method would eliminate that, and is also useable for a building-to-building approach instead of city-to-city.



About your first comment: it's true not all influences have to be recalculated for every passenger. But, when something is calculated doesn't really matter, we only need to ask ourselves on which issues we want to introduce competition between companies into this game. If we only leave company rating as an influence, the competition would be more on the "plant trees and bribe"-front than on the efficiency of our networks.

[Zuu]

..., we only need to ask ourselves on which issues we want to introduce competition between companies into this game. If we only leave company rating as an influence, the competition would be more on the "plant trees and bribe"-front than on the efficiency of our networks.

Good spoken, PJayTycy! You have a an important point.

1) At what frequency should new passengers be created?
=> 1c. should hapen often, as n=in real life, people start thier travels at any day. Plus that spreading out pathfinding should be good for performance.

2) Should passengers be created & consumed by buildings or by cities?
=> 2a

3) Do we assign them their complete route from start => finish or just drop them at a certain station ? (the question on how this route/station is chosen, comes next, but the answer to this question will have a strong influence on it)
=> Leave it for now, and use the fastest variant.

4) What about unreachable destinations ?
=> 4c or 4b.

5) What will influence the passenger when choosing a route.?
=> 5b, 5f(how many changes), 5d



OT - but related:
Last week I have been at practical training on a company that work with Traffic. One thing they did was to create couputer models of reality. In one model they had all streets and public transport links with every single trip. Quite cool. Then they could make calculations with the model to see what hapens if you add a buss line here or change this line here.

To get closer to the topic, the man that demostrated the software and the model for me told me some things about how passengers choose routes.

Diffrent parts of the route is valued diffrently:
Walk to first stop: 2t
Wait at station/stop: 2t
In vehicle: 1t
Change cost: 5 min (At each change 5 min is added to total time)
Walk from last stop: 2t

In addition to these parameters I am sure they use other too, as price.

For each passenger a first selection of routes is made, and then they are weighed, as abouve, to find the route with the lowest cost.

[Steve]

Consider the following network. A passenger at X wants to get to Z.



The algorithm looks at all possible results narrows it down to the following:
XBDZ
XBEZ
XBEHZ
XEZ
XEHZ
XFHZ

All of these routes are deemed to be acceptable. So it waits for the first train to arrive for one of the routes.
Lets assume a train to B arrives first. The passenger boards it and is happily on their way. At B, they narrow down their list of routes to:
XBDZ
XBEZ
XBEHZ
By simply removing the ones that are impossible after that first move.

The passenger then just waits for the first train to satifsy one of the routes. Obviously E has the greatest potential of routes, but that's too much information to process, it's best to just get the first one to E or D.

[Zuu]

I think your aproach is OK.

Locked until the DD discussion arrives at this issue.