Improved airports

[jamespetts]

Interesting discussions, although this is now rather more about scale generally than airports per se!

Firstly, on airports themselves: SDog's screenshot is quite something - it would be most exciting to see something on that scale actually being used for real in an online game. As to the timings within an airport, that should all be realistic enough if aircraft taxi at realistic speeds (which I think that they do - if not, that can easily be fixed). Don't forget that, with every reduction in the meters per tile setting, the game minutes (for travel time purposes) per real life seconds reduces proportionately so that the same number of game meters is covered in the same number of game minutes.

As to map sizes generally, to simulate both local urban transport (which ought not be forgotten in all this discussion about airports - urban rail and 'bus networks with a high scale setting are very fun!) and air/sea transport properly at the same time, large map sizes are indeed required. Large maps are more practical than people seem to think, however. The current Simutrans-Experimental server game has a map with dimensions of 1724x2816 (using 250m/tile, the then default setting) and 675 towns (I think that a player manually founded a new one, as the total is now 676). Once a number of bugs/issues had been fixed in Experimental, performance was perfectly acceptable on that map for all but quite elderly computers even when it got to a very high level of intense development in the 1970s and onwards, and that was with the physics engine as it was in 10.11: Bernd has made the physics calculations (which were previously one of the more demanding elements of the game) much more efficient for the new release.

What really makes a difference to performance is not the raw size of the map (which only affects memory consumption, rather than processor use or memory bandwidth, and memory is not expensive these days), but the intensiveness of the network, which, in turn, depends on the number and size of towns and industries and the number and level of activity of the players. So, doubling the map size whilst retaining the same number of cities will not significantly adversely affect performance except during the initial map generation, which is done once and need not be repeated. Indeed, for an online game, this is done once by the server administrator, so is entirely transparent to the end user in any event.

There are two further tricks for maximising the distance of long distance travel whilst not putting too much demand on peoples' CPUs. One is, as SDog suggests, to have a long and thin map, which is two or even three times as wide as it is high. The other is to have a large area of ocean in the middle of the map, enabling long journeys (by sea or air) but without having much that consumes CPU cycles or memory bandwidth for large areas of the map that have to be traversed.

For example, map no. 1532 (on Experimental) with a West to East dimension of 5,632 tiles (704km) and a North to South dimension of 2048 tiles (256km) and a water level of -1 or even 0 gives a good "ocean in the middle" effect. One could similarly try map no. 68 with dimensions of 5760 x 2048 (720km x 256km) with a water level of -1. (Both have mountain height set to 160 and map roughness to 4). Incidentally, the size in memory shown is the uncompressed size - one need not worry about having to send over a gigabyte down the network: the compressed size is a mere 8.7Mb (and this is with a full complement of trees - I need to thin out the forests a little, I think). Map no. 680 also looks promising with a water level of 0 (this is my favourite so far, I think).

Getting the map generator reliably to produce maps with large central oceans is a rather hit and miss business, so another approach might be to get a very large height map either of an actual region in the world that is appropriately configured or create one with the appropriate shape(s) of land mass(es) to suit the requirement.

Giuseppe - are you sure that you want just Makeobj? No version of Simutrans-Experimental ever released in binary form can read the files generated by the latest -devel version of Makeobj, and the latest -devel version of Simutrans-Experimental is required to read them. I have put both Makeobj and the executable into this .7z file for you.

Edit: With no trees at all, the 5760 x 2048 saved game (with no infrastructure and 16 cities) is reduced to just over 1Mb in compressed size.

[kierongreen]

Great Britain at 125m/tile needs in the region of 8000x4500 tiles (if you rotate it a little - you get some islands on it as well, but not all of Ireland). At least in simutrans standard maps should have less than 16.7m tiles, and at 36m tiles this is too big for standard. This limit is mainly imposed by how labels are stored and rotated I think, so if you exceed it then as long as you are prepared to deal with this it's not a huge problem. You will notice this if you rotate the map, or if you mange to build two stations such that y1*map_width+x1 = y2*map_width+x2. Of course, you could alter the way labels are stored to get round this in experimental.

Playing a map 5000 tiles long isn't a problem, as long as the overall tile count isn't too big, e.g. 5000x1000 should be perfectly playable on an 8 core machine (with standard 4000x3000 or so is just about playable later on in the game on a dual core atom).

Of course, just because physics are by default set to 125m/tile doesn't force you to play at exactly this scale though. You can adjust this value, or just use a smaller map to represent the same area anyway.

Having tested the size of airports myself (and seen carlbakers excellent screenshot) I have to say large airports do look nice though

Are there many flights other than London-Manchester-Edinburgh-Glasgow-Dublin in British islands?

Great Britain to Ireland (most airports in Great Britain to Dublin or Belfast) has a fair few flights and is generally the quickest and most convenient way to travel. London - Scotland is the only major traffic route within Great Britain where air travel has a real advantage, though if you are connecting to onward flights at Heathrow or Gatwick then flights from Manchester, Leeds or Newcastle to London might make sense.

[jamespetts]

I have known people flying from London to Manchester.

Things rather do all have to be to a consistent scale, however, as all the distances between stops, the journey times and the waiting times (and journey times, in turn, are linked to appropriate comfort levels) are all linked into the meters per tile scale as well as the actual physics (stopping/accelerating distances, etc.), so that the relationship between things like dwell time and journey time or airport minimum waiting time and speed, etc., are all fixed to the scale. The whole point of having a larger map is to enable simulation of long distance as well as short distance journeys in the same map.

One way of allowing larger ocean areas would probably be to tinker with the map generation settings so as to alter the scale there: this has already been done once for Experimental, and could without much difficulty, I think, be done again so as to have less fussy maps on that larger scale which would have a greater chance of having a large central expanse of (nearly free, in computational terms) ocean to allow for long distance journeys as well as local transport.

[kierongreen]

I have known people flying from London to Manchester.

Likewise I've known people fly from Manchester to Scotland. Over these sort of distances train travel makes more sense really though, unless you happen to be near the airport in both places. Mind you, most people would probably drive these days anyway :(

[Carl]

Likewise I've known people fly from Manchester to Scotland. Over these sort of distances train travel makes more sense really though, unless you happen to be near the airport in both places. Mind you, most people would probably drive these days anyway :(

Depends where in Scotland! Manchester -> Aberdeen is 1.25 hours by plane, but can be 7 hours by train! And often similar costs...
(sorry, off-topic!)

[kierongreen]

This was Manchester to Glasgow! If you are travelling from anywhere north of the central belt in Scotland to England you are usually better off flying in terms of time, assuming there are flights available. Also flights from Aberdeen (I'm thinking Eastern Airways here) can be very expensive

[prissi]

Also flights shorter than 150 km are very common on islands, like in the caribbean (for instance Purto Rico to American Virgin Islands), or between the Hawaii islands. One of those company used really vintage 737, which were still not logging enough flight hours to be disposed, because they only flight less than 40 min for each leg.

Wikpedia say: "Aloha Airline's longest inter-island route was 216 miles, while the shortest route was a mere 62 miles. Average travel distance per inter-island flight was 133 miles."

[Milko]

Also flights shorter than 150 km are very common on islands, like in the caribbean (for instance Purto Rico to American Virgin Islands), or between the Hawaii islands. One of those company used really vintage 737, which were still not logging enough flight hours to be disposed, because they only flight less than 40 min for each leg.
Wikpedia say: "Aloha Airline's longest inter-island route was 216 miles, while the shortest route was a mere 62 miles. Average travel distance per inter-island flight was 133 miles."

Finding (google) "aloha airlines flight 243", contains a description of a plane crash due to excessive number of departures per day that the aircraft suffered. Too many daily flights of short duration ...

@James
The airplanes will rise from the ground after covering the tile set in the file ".dat" or the aircraft will walk accross the runway?
The acceleration of the aircraft will be phased in starting and take-off or a good part of the track will be flown at a constant speed?

Giuseppe

[jamespetts]

Giuseppe,

I have not done anything with aircraft physics yet, but I do plan to make what looks like it will be a simple modification of making the aircraft's rotate point when taking off the same as its minimum runway length, so that players are not baffled when they are forced to lay a long runway, but aircraft only appear to use a small part of it.

Edit: I have been doing some tinkering on the -devel branch - map no. 2147483642 works quite well with the new settings.

[Milko]

Hello

The new features are:
(1) aircraft all have a minimum runway length in meters, and cannot take off/land from a runway shorter than that length; and
(2) all airports must have control towers.

Next step: Airplane Range?

Giuseppe

[jamespetts]

I have considered range; however, it is a tricky issue. It would apply not just to aircraft, but rail and road vehicles, too, as well as ships, and, indeed, almost any non-electrified vehicle. The trouble is that it is not an easy thing to communicate to a player: whilst actually stating the range in the depot/replacer window is easy enough, players actually being able to work out whether any route will be in range of suitable vehicles before planning it is not an easy task. I suspect that people would find this feature a bit too awkward and fiddly.

I might take a different view if there seems to be overhwelming support for it, but it would also be much more tricky to programme than runway length restrictions, and would fall to take a lower priority to a number of important outstanding issues such as increases in vehicle maintenance with usage, overhauls, way upgrading costs, and so forth.

[kierongreen]

Also on the scale of simutrans maps range isn't an issue as it would be larger than the map...

[sdog]

it would be also a bit difficult to decide what to do with a plane, when it runs out of fuel.

limiting the routefinding before it starts is also not an easy option, since waypoints might be very exploitable in this regard. Short flights and endless circling would also not be covered.

I don't think it would be uncalled for if only planes would be limited in range by fuel.

It is a non issue for ships, you can just pack enough to reach any destination port. Road vehicles could just fill up unnoticed, as could diesel or steam trains. The infrastructure required is minimal.

[Vladki]

refilling of trucks and trains, could be done as special station extension. It was this way in railroad tycoon, whre you had to build water tower and sand tower at least in hub stations. If you didnt your trains ran out of steam and sand and moved veeeery slooowly.

[ӔO]

One thing that might be worth consideration: variable running costs with acceleration/deceleration and constant velocity.

[el_slapper]

Range is really somthing that would kill the gameplay for not much impact. Especially on Experimental where the scale is small, & the biggest you can simulate is the British Islands. Truck stops & refuels, other vehicles have far more autonomy than the size of the map.

having refueling stations would horribly clutter the map with little impact on realism or gameplay experience. I vote no. Interesting problems are worth solving, like infrastructure size depending on the vehicles used - and that's exactly the purpose of long runways.

Railroad Tycoon had far less things on the maps, and far less things to do for the player. Lay tracks, add refueling stations, run the train. Even like that, it was rather boring to add refueling stations in the middle of the sahara. You are dead if you don't add them, but there is not much strategy in adding them : you put them when the loco runs out of steam/sand. Not an interesting decision. Choosing wether I will use airplanes or trains between London & Manchester, and building according airport/railroad IS interesting, OTOH.

[isidoro]

Well, there was some strategy involved in placing sand/water stops since they cost money.  Similar to placing depots in ST.

[sdog]

It is a little different for planes though, as it might restrict the choice of some planes for some conexions.

What would be the shortest range any of the planes included or planed for the pak has?

The longest distance would be less than 1000 km. Going diagonal on a 6000 tile square map (6k*1.4=8k) would provide that, or a 8000 tile elongated map. That is quite a lot at this time.

A Cesna 172 has about 1000 km range. That's about as small one could get with a modern aeroplane.

It might matter for historic planes, early Hadley-Page 'planes were flying Croydon to LeBourget, 350 km. I should guess they were designed specifically for this and it might be about the maximum possible. On the other hand, the customer would have changed the requirements, so the range always would have covered the distance required.

[o_O]

No matter how big the map gets, realistically most large aircraft traffic would still be to and from off map locations.  It might make sense to add abstract international destinations to allow realistic airport sizes, passenger levels and transportation requirements while only modeling a smallish region.  So you schedule your Airbus A380 to fly to Shanghai twice a month, but really it just flies off the map then re-appears a while later with new people who want to go someplace.  This way you could have huge, busy airport hubs without specially designing gigantic maps or fudging the airplane numbers.

Although I don't know how well the code could accommodate this. 

[jamespetts]

o_O,

actually, I have been considering this idea recently. The way that it might work, I think, is to have special tiles that, when set as a destination, allow a ship or aircraft to travel a designated distance off map, load and unload passengers/mail/cargo of specified types in a fixed time, and return to the point again in a calculated time. The vehicle itself would calculate its speed and loading/unloading time as normal, and the distant port/airport would be represented as an actual stop, but the actual distance between it and the special tile would not be represented by part of the actual game map. These special tiles could be highlighted red (for example) and be restricted to being water tiles on the map edge. Players could see all vehicles in transit on the relevant overseas route by clicking the tile, and would be able to see the convoys as normal, just with a black background instead of an actual slice of the map in their pictorial window. This would indeed allow simulation of international transport within sensibly sized maps.

This should be reasonably possible to accommodate in the code, but would still probably be a rather large amount of work. It is not currently a priority for my coding work, as I am concentrating on balancing (and possibly merging the latest Standard features) for the release after next (the next release being a matter of bug fixing and testing) of Experimental, but this is something to bear in mind, I think, for one day in the future.

However, if anybody else would like to code this, that would considerably expedite the process.

[ӔO]

one could just paint up a complete airport as an attraction.

[jamespetts]

Yes, that's another method that I've considered in the past. Both have their advantages. The advantages of the airport as attraction method are that it takes far less work to code (although actually producing a graphic for a full sized airport would be quite a challenge).

The main advantage of the overseas transport tile method is that it allows actual simulation of long distance travel, which is not currently possible in Simutrans (practically), and is an interesting thing to simulate, especially with a good range of aircraft. It's really a matter of balancing one against the other.

[isidoro]

Conceptually speaking, what is the difference between such special tiles and industry or attraction stations with an increased loading time (simulating the round trip off the map)?

I say that because maybe that is a not so difficult way of programming it...

[jamespetts]

Conceptually, the difference is that, with the special tiles, players actually have to manage and control international transport. With attractions/industries coded as ports/airports, international transport is assumed, rather than simulated.

[Carl]

Yes, that's another method that I've considered in the past. Both have their advantages. The advantages of the airport as attraction method are that it takes far less work to code (although actually producing a graphic for a full sized airport would be quite a challenge).

The main advantage of the overseas transport tile method is that it allows actual simulation of long distance travel, which is not currently possible in Simutrans (practically), and is an interesting thing to simulate, especially with a good range of aircraft. It's really a matter of balancing one against the other.

I don't see these methods as mutually exclusive.

[ӔO]

personally, I don't see a need to allow players control over international, off map, transport.

[sdog]

The airport attraction method is too much of a hack. Possibly unbalancing the attraction system, by having one mega-attraction. Also if there was only one, a single player could effectively claim it in multiplayer, gaining a massive advantage.

Have you asked simutrans standard, if there was a desire to have a long-range connection point on the edge of the map? I wouldn't be too surprised if there was a strong desire, (and a steep potential barrier of opposition to overcome) but it might be sensible to ask at least. Some brainstorming on how this would have to be implemented, such a sink/source needs proper self balancing attributes, might also be helpful.

[Carl]

The problems sdog raises don't seem insurmountable. The passenger level could easily be balanced relative to other attractions in the pakset. And in network games there could be a rule that airport stops/stations must be public.

[jamespetts]

Whilst the two methods are not technically incompatible, it seems to me that it would be very odd and confusing for players if two very different ways of representing the same thing were used simultaneously in the game; I don't think that that would be desirable at all. It would be fun indeed to simulate long distance overseas transport, but the question is one of practicalities.

Aside from the issues as to the coding of it, a further thought has sprung to mind as to a potential difficulty: in the days of sailing ships, it could often take many months to reach a destination. On the basis that there are 720 hours in a month, a four month journey would last 2880 hours. On the basis that there are just under three and a quarter travelling hours in every Simutrans game month, the journey lasting four months in real world time would take just over 886 years in compressed Simutrans time, which would clearly not work. For aircraft with journey times of not more than 24 hours, this would be less of a problem, but even a 20 hour journey would last over six game months in Simutrans-Experimental.

[Carl]

Whilst the two methods are not technically incompatible, it seems to me that it would be very odd and confusing for players if two very different ways of representing the same thing were used simultaneously in the game; I don't think that that would be desirable at all.

One need not think of them as representing the same thing. Most pertinently, airports generate high levels of retail and service employment, which generates a lot of traffic to and from airports. As such we should not expect travel to airports to be exhausted by those who are getting on planes. If simulated off-map air travel were to be introduced, one could think of the "attraction" role of airports to be limited to traffic of airport staff, etc.

[jamespetts]

Wouldn't that be even more confusing, as players would expect that people would go to international airports for both reasons? Indeed, it's rare for people to go to airports if they don't actually want to use the airports as airports - the shops and things are generally no better than on a high street and are intended for and used by air travellers.

[Carl]

I was mainly talking about employees, not shoppers. Employment levels at and around airports are such that there are often transport links which are put on especially for the purpose of getting workers there. Even commercial links owe a significant portion of their ridership to airport employees. Whenever I've used the East Midlands Airport buslink from Nottingham, for instance, a significant amount of passengers were airport workers with season passes rather than travellers with suitcases.

An example: a quick search reveals that Stansted employs over 10,000 people. That's a lot of non-flying traffic.

[jamespetts]

Ahh, I see what you mean about employees. Still, I do think that it'd be confusing and counter-intuitive to have airports as attractions just for their employment rather than their international passengers.

[jamespetts]

The airport improvements discussed herein are now in version 10.12 of Simutrans-Experimental, released to-day.