NAM Traffic Simulator Development and Theory

[z]

Could you elaborate? How did you go from "11 billion sim years shows this is the best we've come up with" to "Nope, the older one's better"?

There are a couple of reasons here.  First of all, deriving these numbers is a very complex process, and there's a lot of noise in the data.  This is the reason for the 11 billion Sim years of tests.  One of the reasons for the reversion is that I decided to put the traffic simulator through different, additional types of stress testing.  These tests produced bigger differences than the earlier tests, specifically in the area of abandonment due to commute time, and they all favored the original PH.  But then, of course, there's the question of why the newer PH did better in the original tests.  We could even see that on some occasions we were getting better paths, and we weren't seeing any worse ones.  However, as we repeated these tests more times, we kept getting different results, and after the additional rounds of testing, it was clear that the PH wasn't making a difference in the pathfinding.  Further research at the MTA turned up the surprising fact that there are actually two distinct parts to the pathfinder - the part that navigates through transit stations, and the part that navigates everywhere else - and they operate very differently.  Only the second part is affected at all by the PH.  However, by doing more original research into transit switching, we were able to come up with transit switch points that provided much better results from the pathfinder, which translates into better performing stations.  The first batch of these upgraded stations were included in NAM 32.  At the same time, we were unable to obtain perfect results with the transit switches for multi-network stations, and it appears from our work that perfect results for such stations are not possible in this game, regardless of the PH.  However, the second part of the pathfinder, which does not deal with transit stations, appears to be producing perfect paths all the time at its current value of .005797.

I hope this answers your question satisfactorily.

[HappyDays]

Intriguing. One question down, three more ready for launch.

1: MTA = ?

2: What is it about the transit switching system for transit stations that affects pathfinding?

3: Are there plans to update Road Top Mass Transit (Currently at v3.6) to meet the new transit switching standards?

[z]

1: MTA = ?

It's the Mass Transit Authority, a wholly owned subsidiary of The NAM Corporation^TM.  The MTA has two subsidiaries:  the MTA (this is a different one; it's a semi-recursive acronym for "MTA Turnpike Authority"), which is responsible for tollbooths, and Nocturnal Aviation, which is responsible for the airports.

2: What is it about the transit switching system for transit stations that affects pathfinding?

Suppose the pathfinder comes across Brenda Xne's monorail station in its search.  This station offers bus, subway, and monorail service.  Which one of these should the pathfinder take, if any?  Ideally, it should be the one that produces the fastest route.  But this is the first part of the pathfinder working, and it doesn't navigate transit switches very well.  So unlike the second part of the pathfinder, it has problems picking the travel type that will result in the fastest route.  By picking the right transit switches to include in the transit switch point, we are able to help the traffic simulator find the fastest route.  We are well along the way of nailing down the final details of which transit switches work best in which types of stations.

3: Are there plans to update Road Top Mass Transit (Currently at v3.6) to meet the new transit switching standards?

Yes; all of the new transit switches will be present in RTMT 4.0.  I have been copying the transit switch points into RTMT as they have been completed.