Programs for automatization routes building

kbodjar

New member
I would like to hear the opinion of programmers regarding the means for automatically creating real routes. Well, at least landscapes, elements of nature.
So, my thoughts on this matter:


1. Creation of relief, basemaps - TransDEM is almost certainly ideal for this. One automation tool already exists.
2. Filling the map with the environment: here you need to implement some kind of system of object snapping to coordinates so that the right tree is in the right place. What can be in the role of anchor? Colour? For example, if you make basemaps in TransDEM, then in Photoshop, paint visually defined areas with colors corresponding to a specific class of objects (shrubs, conifers, mixed, etc.). Next, in TransDEM, select the painting of the ground according to the basemap. Then, through a script, place the object on all the places painted with this color. Rotation randomly, accuracy is ideally adjustable.
For buildings it is more difficult: the rotation angle must be taken into account. You can not take into account, rotate after placing manually. But in this case, you will need a bunch of building classes: sorting by number of storeys / newness / purpose. Therefore, it is necessary to create a kuid database, as I understand it, which will be loaded by the script.
With splines it is still more difficult, most likely only by hand. Although in the same TransDEM it is possible to export paths directly from the program to the map. Of course, without considering the track profile. To take into account the profile of the path, you need to determine a certain zero point, set its height, and then dance from this point, using the profile of the path, again, from some table (a database in which there is a distance of the section and its slope). The maximum that can be done in this way is spline embankments, roads, grass. And for splines, a classification must then be made too.
3. The ultimate dream, of course, is the creation of such things based on AI, which would analyze materials, such as video from the cockpit, and build a landscape. But this is absolutely fantastic.


So, what are the chances of doing something like that? Basically, to collect all the automation methods invented by people into something coherent. The construction of routes will accelerate, the number of abandoned and long-term routes will decrease, I think this will only make everyone better.
 
I mentioned something similar to this in relation to the new Microsoft Flight Simulator and how it draws the world.

Sadly I got the vibe that 1. There is little interest among route builders who prefer to plant every tree and blade of grass by hand and, 2. Implementing the type of world database and linked photogrammetry seen in MSFS would be quite beyond the resources and capability of a small company like N3V.

Our best hope is that (secretly) Asobo, Blackshark and MS are looking at how they can adapt their game engine to ground transportation including railway operation.
 
Generating terrain, buildings, and roads from Open Street Map data is not perfect yet. I have an add-on for Unity3D software that generates terrain, buildings, and roads, and I am not happy with how the add-on generates terrain, buildings, and roads.
 
Each of us has a different set of interests when it comes to using Trainz.

In my case I am one of those route builders who prefers to "plant every tree and blade of grass by hand" but other Trainz users only ever download pre-built routes from the DLS and DLC and drive trains. I have no problems with that.

My current project, which I started when TS2004 was the latest version of Trainz, is now nearing fruition in TRS19 about 16 years later. I would hate to think that some pimple-faced adolescent, barely out of diapers, could click on a few buttons and recreate my entire route, all 375kms of it, in just a few seconds. Where is the fun, where is the sweat, where are the tears in that? Michelangelo suffered in his labours, as do I :'(

Such a vision could eventually come to pass but it would have many problems to overcome. For example, how would such a system even know that "the right tree is in the right place"? I fear that my South Australian narrow gauge desert route would be populated with forests of Nth American Spruce trees (or some "trans-Atlantic genetic hybrid"), that it would be signalled with the latest USA technology instead of the lower quadrant semaphores of the steam era, the tracks would be ballasted to take 200t ore wagons instead of with fly ash or dirt that would carry just simple 4 wheel 5 plank wagons.

Enough of this "technology rules" nonsense.
 
I'm another prefer to do my own work and create the correct buildings and such if not available and then there is the historical and geographical accuracy aspect. Be Impossible to get a say 1930s or 1890s video and auto create buildings that no longer exist from it, if that was even possible without a super computer.

Most people make routes initially for themselves and may or may not choose to share them and it can take years of constructing and research to get anything that is anywhere near accurate. Can't see automatic placement of stock assets achieving that.

Believe it or not most Route Creators enjoy that part of their hobby, if they did not they would hardly be creating routes!
 
Thanks for the answer. Of course, creating a route is a hobby. But there are people who just want to embody their native spaces in the Trainz environment, to ride on them. They begin construction, but doing everything manually is laborious, and often abandoned. Therefore, the speed of construction and the help of a computer in creating landscapes of the same type are important. In any case, this is the basis, followed by manual revision of all details, if necessary.
 
There is a place for both, IMHO as someone who builds or at least has built, both prototype and model/fictional inspired routes.

However it can just be totally daunting to look at 50 or 60 miles which requires comprehensive scenery placement when the technology exists (at least elsewhere) to at least populate 90% of this in a few moments.

And obviously yes there have been issues with how the autogen renders things in MSFS. The patch before last had office blocks and flats poping up in rural villages, subsequently fixed. In the case of a route, the autogen would be to produce a one time overall coverage of the route based on the mapping and photogrammetry, which the author could then go in and detail at their pleasure.
 
I think there are some good ideas here. What if you could start with a large terrain only route from TransDEM, and then using a separate add-on program, sold by N3V or under license by them, you would be able to see contour lines and play with different options to develop baseline climate zones, texturing and vegetation. This add-on program would have enhanced terrain tools from what is available now. You would still lay down track and road splines and whatnot yourself because most route builders would prefer to do splines and buildings on their own. The result would be large areas of background terrain that the user designed. Then you would save, exit the add-on program, and start Trainz for route building as usual. Everything generated by the add-on program could be modified within Trainz so the essence of route building would not change from the present.
 
The problem lies in the data available. There is a thread right now active that I'm involved in about how accurate you can get a DEM to look in Trainz. Quite a bit comes down to the type of DEM's available and it's not available everywhere including all of the United States. One example where I am Buffalo, NY across the border from me does not have 1/9th where we are stuck with 1/3 or below in some parts right across the border in Canada where I live.

Even when you have 1/9 you would think it would pick up a double track roadbed 10m wide but I haven't seen much evidence of this.

So once you generate a route just in TransDem as amazing the scenery looks around it where you may not notice the details, you will see them in the track itself that rides like a roller coaster even on a 5m grid. Even patching the areas that are normally culverts, creeks, overpasses, and rivers, the track is still not at the actual level and I have 1st hand observations of this.

This isn't the fault of TransDem. I'm a programmer and I would touch a project like this. I've done car forwarding software for Run8, and I'm reluctant to do anything for Trainz at this time programming wise because of all the variables of car types, etc. I honestly don't see it happening.

I'm all for options as well. Give the options for those that want the details, to those that don't. The issue here is either option right now to me is not plausible. But nothing wrong with asking.

Thanks

Sean
 
I still think that the idea can be implemented in the form of a program/game script that works on the principle of replacing the earth texture with objects that the user would specify with a list. The earth texture has a kuid. For example, when the program starts, it looks for whether kuid's, from the built-in list, is on the map. By Convention, the texture fills the area for objects. The meaning is the same as in auto-generation in Trainz 2019, but it must be provided not only for rectangular selection zones, but for arbitrary ones, or with reference to the earth texture! Unfortunately, I'm not a programmer, but I can't believe this is hard to implement.
 
how would such a system even know that "the right tree is in the right place"? I fear that my South Australian narrow gauge desert route would be populated with forests of Nth American Spruce trees (or some "trans-Atlantic genetic hybrid"), that it would be signalled with the latest USA
The idea is to create an initial SKETCH of the landscape, not trying to depict as much detail as possible, although that would be better. According to the idea, you can make any adjustments at any time, and only you determine the list of the same trees that you want to see in a certain area.
 
I think that OSM data isn't perfect. For example, town houses are built as several houses combined into one building. OSM data thinks that town houses are separate houses, not combined into one building. Some buildings have wrong kinds of roofs. If you want a perfect model of a city, you'll have to get into your car and take pictures of every buildings on every street. I think that Google Maps did this using cars with special cameras on them.
 
I think that OSM data isn't perfect. For example, town houses are built as several houses combined into one building. OSM data thinks that town houses are separate houses, not combined into one building. Some buildings have wrong kinds of roofs. If you want a perfect model of a city, you'll have to get into your car and take pictures of every buildings on every street. I think that Google Maps did this using cars with special cameras on them.
Nobody talks about using OSM. Classes can be set manually, similar to the HOG program, where a line with color 255.255.0 was used for the path filter, which was drawn in Photoshop
 
I must be maturing. I can see both sides of this discussion.

If automated route creation was possible, it certainly would not be mandatory, so I can't see why anyone would oppose it in principle.
 
I don't oppose any of this (not to saying the last post says otherwise to myself). The OP can say that they're not talking about OSM. I can assure right now that even the best DEM out there 9 times out of 10 will need adjustments. In fact I recall the author of TransDem who is a programmer much more qualified than myself mentioned OSM as overkill in a tutorial on allowing all kinds of vectors to be placed by TransDem and other programs. In other words it can already be done to a certain extent.

I'm not trying to be argumentative, but realistic when it comes to what is possible. OSM is the closest thing out there that will pick up vectors used to place objects. This is why people are mentioning it and I conquer. An images color can be anything from say a soccer ball to a sprinkler (insert whatever else). I use level 19 Bing images (20 is only available in bigger city centers) and I can barely see the tracks, let alone bridge piers and bridge details. Most of the assets we have never match a project and they often need to fit (how would a program choose this?). I walked .5 miles of track today to get an up close look at a creek bridge for example and found what kind of bridge it was, what kind of piers, and it's height which the 10M DEM didn't register. Even then I couldn't find the perfect bridge, or piers. I eyeballed it at about 15 feet high from ground to track with about 12 feet of clearance underneath.

I'm afraid this is a pipe dream. Do I think in about 10 years will we have the kind of data necessary to produce automatically absolutely. From a flight simulator standpoint it's much more plausible because if your flying you won't see all these details.

Now if there is a programmer willing to take on such a challenge more power to everyone. I believe in options and nobody is going to force anyone into this automation (anyone thinking that are worrying too much). I think the examples of people saying they would hate to see it on their developments is if they used it not if they were forced to use it.

Again if it can be done, great. But for me personally I'm about data and facts. If this is about automatically building a route, what kind of data would need to be built up (and with what program?) that would be simpler than the route editor and it's own imperfections.

Thanks

Sean
 
Last edited:
I too can see this be an optional thing for those that want, let's say, a randomly generated route based on region. The tool could set the number of baseboards, aka size, hill height, number of cities, towns, and industries. This isn't much different than those tools found in Transport Fever 2 and Transport Tycoon among other programs of similar ilk.

Random city generators I can see useful for background cities, but not for those that are up close to the tracks. There are some such tools available for the architectural and city planning industry. This plug-in integrates with Arc-GIS and other programs to generate buildings base on specific parameters and even maps. While they look really nice from a distance, the output is pretty poor with building that look the same no matter where in the world the city takes place.

Using displacement maps, we can already generate terrain, albeit, in a more laborious process. There are various websites and programs that will generate fractal landscapes. The output is a grayscale image that is then used as a displacement map. These are used for various programs and games such as Cities Skylines, Sim City 4, and other building-type games.

Track while convenient maybe a tricky one to implement. The user may have to connect the cities on the map prior to generating the terrain, or perhaps afterwards, before the output is finalized.

For those of us that like to manually create our world, the one thing that I think would be most helpful. A yard generator. Yes, something that would create our goods yards easily. Creating a yard is the most difficult thing I think and very frustrating too We have numerous templates, some old fixed-track assets that no longer work, and other things that assist us in creating a yard, but in the end our track is crooked and messy. Having a tool to create the these using set parameters, would be most helpful.
 
You mean lack of. ;)


I suspect an element of copyright / IP is involved or as the board format was changed to accommodate PBR ground textures, it's possibly not finalised yet.

Personally if it ain't broke, comes to mind. ;o)
 
Back
Top