Astro Trains: Lets Talk

[Draco18s]

From another thread:

Spoiler for Hiden:

they gave AIP on death (which [...] is my main beef with Astro Trains)

Let's talk about Astro Trains here for a minute, shall we? After all, one more *derailed* topic won't change the face of the forum, right?

My opinion is that it needs much more player agency. I mean that players must be empowered with a clear mean of altering the state of the game. Right now, trains' routing is barely predictable and players have nearly no control over them. I would be okay with the AIP cost if I could clearly see the result before. In other words, I want to see what I'm paying for. I remember the idea to make trains only travel from station to station if they are on adjacent planets. That would make understanding of train routing crystal-clear, and the players' action would be "cut the rails". Then an AIP cost to force the player to cut only the interesting rails would be acceptable. Also, cutting a graph is already something interesting to make with Special Forces and planet capture (and Hunter plot to spice it up).

So I had a drug/alcohol/sleep/deprived idea last night about astro trains and the main thing that makes them such a problem (and the above post reminded me):
The inability to see their routes.
With a slight tweak, we can fix this:

During map-gen, each station precalculates its routes to each nearby station (not ALL stations, just the nearest three or four) and lays down "train tracks" that are essentially non-interactive visual display of train routes.  It could be as simple as something like the lane wisps in Dungeon Defenders 2.  Immediately the player not only knows how many routes pass through their systems, but where to find the stations ("just follow the yellow brick road").  Every train would use this micro-network to path through the galaxy. If it wants to go from A to D and there isn't a link from A to D directly, then it must pass through Stations B and C first.

Upon destroying a station, all of its links would be nulled and the network would recalculate in order to maintain a connected network graph, adding new links anywhere in order to keep the overall connectivity high.  Yes, this may result in new routes through the player's territory that didn't exist previously.

What this change does is make the trains more predictable.  The player can build around it ("There's a rail here, maybe I should build my command station over here instead"), have agency in fighting back ("I'm hunting wabbits!"), as well as restrict where trains can go.

Second: Destroying a train means something

One of the other things I recall about Trains is that if one of them is destroyed, the AI network just creates another one out of thin air to replace it, to keep up some ideal #of trains value.  Trains being as monumentally indestructible (for the most part) as they are, it would seem that they should take a fair few resources to create, and while the AI does, essentially, have "infinite resources" and just isn't devoting them to this part of the galaxy, it similarly shouldn't care about one train going poof so much that it instantly replaces it.

Give these things a build time, force the AI to wait out that timer.

I'm ambivalent about trains that reach their destinations. I'm ok with them disappearing into the ether and popping out somewhere else (as something else), as that can be handwaved as having been reformatted, put into storage, or whatever that we can't see going on off-screen in the abstraction.  That said, I do like the idea of a cargo train having to request a combat train travel to its origin point before it can set off (giving the player additional opportunities to slow down those projects: if they intercept a combat train and destroy it, the cargo train needs to request a new one and wait longer).

Third: Train Goals

Trains need to have a goal beyond "move from point A to point B as arbitrary train type rand(9)."

I think they do currently, or at least the cargo trains do. In that every one that arrives as its destination adds resources to some AI super project (super fort, etc.) on the AI homeworld.  I want to see this expanded and fleshed out a bit more.  "Reach its destination" is really super vague and doesn't mean a whole lot if you reduce the train network to 2 nodes that are in adjacent systems: now not only is it easier for the AI to complete those transits, but it's harder for the player to intercept!  If this link was always that vital to resource movement, why wasn't this route being utilized all the time?
(This is the main reason why having trains on a "low intensity" setting is actually worse for the player than a higher one!)

 * Step 1: create virtual "resource nodes" adjacent to some train stations (far from the homeworlds) where the cargo trains originate (these nodes may not necessarily be visible to the player, and indestructible if they are).  These trains must travel from these stations to "a goal station" in order for the resource counter to be incremented for the super project.  Additionally the AI homeworlds would start the game with a station; these stations are only goal stations unless no resource stations exist, then they are considered resource stations and not goals.
 * Step 2: create actual project destinations.  The super project will be built in this system or an adjacent system.  Project destination stations must not be 1 link away from a resource station when created (station destruction and rerouting may result in single-link trips, but any new project must respect the two-link minimum, if no two-link+ routes exist (fully connected network) then the project must be built on a AI homeworld and the resource node must travel the longest galactic distance).
 * Step 3: Projects: One project may be a "train station constructor."  After being built it flies to a system that has no train station in or adjacent and is not a dead-end system.  Any links that travel through this system are removed and both end points now link to the new station.  After that, establish a minimum of 1 new link to the nearest stations, bringing the total links up to a minimum of 4.  In the case of a resource node (step 1) no longer having a station connected, these station constructors would prioritize re-establishing that station.  In the case of no resource nodes available for projects, the cargo trains originate at the AI homeworlds, simulating the AI diverting resources from other exo-galactic projects.
 * Step 4: Cargo trains are never unaccompanied. They always travel with at least 1 "combat" oriented train as escort.  What type and how many would depend on AIP, plot intensity, and AI difficulty.  However, on-death, cargo trains drop 100k worth of resources for the player, with another amount added to the local debris field.  If the escort trains are still alive, they may raise the AIP for the death of their escort target (dependent on train type; there could be variations, such as a scorched earth train that nukes the planet when its escort target dies).  Ostensibly this implies that combat trains are fightable.

• New "combat" oriented train: Protector
• Tough, durable, and heavily armored itself, it shields its escort from damage.
• Carries a remote shield generator, which projects a armored force field bubble around its escort target (any other train, prioritizing cargo trains)

[x4000]

This is FASCINATING.  And doing things like showing the train routes are one of those things that we absolutely can do now that we're using more 3D tools.  I really like the ideas here, because the AIs having more basic "verbs" to do things is important, I think.

I don't really want astro trains to be one of those random secondary things any more than special forces guard posts or CPAs are.  Giving the AI more things that they are up to and that matter to them and you is important, and supply lines are a key part of that no matter what.

Whether or not that really fits the mold of astro trains as we traditionally think of them I don't know, but at any rate this is very interesting as a general thought process.

[Steelpoint]

This whole topic reminds me a bit of the raids in Xcom 2.

Its clear the AI is up to no good somewhere else that requires a lot of resources, maybe look to making Astro Trains carrying a lot of supplies (Metal and maybe portable energy generators/shards) that the player can attempt to raid and steal for themselves.

The advantage being you get access to a lot of resources, but you're going to shoot up your AIP since you're depriving the AI of crucial resources.

Otherwise I like this suggestion, simply because I feel the current Astro Trains are just a annoyance for the sake of one, with no real purpose beyond that.

[x4000]

That's a really good idea.

The astro trains were originally invincible, by the way, in their first form.  They were set up in such a way that they had a couple of purposes:

1. Depending on their paths, there were certain wormholes that were guaranteed to take certain amounts of periodic attrition and thus not be completely defensible.  The goal here being the differentiation of wormholes.

2. Depending on where stations are, they could encourage raids into deep enemy territory that otherwise you'd have no reason to do (back then).

3. The stations were destructible but had an AIP cost, so finding them and choosing to destroy them could let you shore up your defenses at XYZ wormhole(s), but at an AIP cost.  The goal here being yet another decision point, and another thing that makes planets more unique.

This morphed a LOT over time, and the trains got more and more annoying.  Arguably they were annoying to start with, but they had more of a purpose at least.

[Draco18s]

Having a couple of types of train that are outright indestructible is fine, provided that it serves a purpose (other than to poke the player in the eye with a sharp stick and yell, "Neyh neyh neyh neyh! You can't touch me!")

The cookie golem is one such unit indestructible unit that serves a purpose: do not be in this system.

The problem with the original system (where all trains were indestructible) was that the player did not know why the trains were pathing through their system.

Oh.  New thought.

Hack The System

If some means of hacking the train network to influence the graph (e.g. to remove a particular link to route trains around your systems without destroying a station) or the station constructors (to influence where it builds a new station) would be very cool.

[Aklyon]

More hacker uses is always an interesting choice.

[tadrinth]

I like the idea of clearly-communicated fixed routes, but I don't think the network should be randomly recalculated when you blow up a station.  The result of blowing up a station should be predictable, so you don't blow up a station and discover you just made things worse.

I think separating 'trains that make holding particular planets very difficult' from 'trains that you need to kill' might help.  There's no good solution in the current design; if you route them away, you can't kill the cargo trains.  If you route them through your planets, then you get wrecked.  In theory you could maybe route them through a planet you don't control but beachhead, to kill just the cargo trains.  Not sure how well that works.

What we really need is the ability to build our own Astro Train plots using the mod system.  A high bar, but a good one to hit. 

[Pumpkin]

I'm very interested in the trains becoming part of a vanilla game (players still able to disable them, like CSG or Strategic Reserve with Lazy AI but ON by default). They sure have a role to fill (like the Hybrids, but just like them, they must move somewhere else).

+1

[Draco18s]

I like the idea of clearly-communicated fixed routes, but I don't think the network should be randomly recalculated when you blow up a station.  The result of blowing up a station should be predictable, so you don't blow up a station and discover you just made things worse.

Fair point that the result of blowing up a station should be predictable.  And I didn't really mean to say "this will be random" but more of a "it follows these set of rules:..." even if it wasn't entirely transparent to the player.

The point of adding the new links was so that the train network didn't become fragmented: that is, there must always exist a route between any two stations.  Kind of how the wandering rogue planets can be smashed into other planets and remove it entirely from the game and the special case of "if this cuts off an AI homeworld..." rule of "oh god oh god, I do not want to do this" action, except that in the case of the train network, it shouldn't blow up the entire galaxy.

[Toranth]

Hack The System

If some means of hacking the train network to influence the graph (e.g. to remove a particular link to route trains around your systems without destroying a station) or the station constructors (to influence where it builds a new station) would be very cool.

Close upon this is one of the ideas I've always wanted:  Make Train Stations into capturables.  Then humans can build their own trains!  Resources cars that travel between human stations, producing additional Metal as they move (through AI space).  Combat cars of various sorts, that do the shooty-shooty bit.  Shield cars.  Transport cars.
Add in the ability to direct your trains to AI stations for more fun.  "Here, AI, special delivery for you!" as you unload the Transport car full of Mk IV ships...

[PokerChen]

Okay, after reading Draco18s good opening suggestions, I'd like to present my alternative vision of how Astro Trains fit into a sequel. This combines some of my thought and existing contributions so far:

Design Motivation
Astro trains represent the logistic network and supply routes of the AI (and potentially players), as the equivalent of trade convoys in an ultimate command economy. This functions as the AI's source of economy inthe local galaxy, and reduces the magic auto-reinforcements from intergalactic sources (which should be reserved for large counter attacks like CPAs and Exo-Strikes).
Two tiers of Astro Trains exist:
- the 90% traffic that represents AI's adoption of the human's old logistic systems, shipping resources from mining to factory worlds. THis is intended to supplement local pressure.
- the 10% special convoys that are usually required to complete AI Plots and certain Minor Faction Plots.
These provide the context and encourage for players to hijack or sabotage train networks.

Layout
Train stations will be located at:
- rich resources worlds (the AI utilises the same metals as humans).
- fabricator worlds (from which AI draws additional local "man"power). This is also where replacement trains are created.
- additionally, stations wil be created at project worlds (secret locations to carry out Plots, can be decoy).

The running network will take a reactive approach:
- Trains will run back and forth between source and destination.
- The length of these deliveries will be large for special projects (cross-galaxy delivery of necessary rare materials), but relatively short for normal traffic.
- Successful deliveries usually means the same route is taken again. Well-connected galaxies should also have multiple routing (e.g. all-6 jump routes should have equal weighting, so that not all trains doing that mission travel through the same nodes).
- When a player destroys a convoy, other trains travelling through that system has a decent chance to avoid that route.
- When a player keeps destorying convoys, trains decrease their cargo carriages and increase theit armed carriages.
- When desperate, AI can divert fabricator trains towards special project.

All convoy are defended by AI fleets, which constitue the majority of defenses. This also allows AI Train booster to be effective.

AI goals
Standard AI Fabricator worlds will produce several things: Special forces, AI Fleet, CPA, etc. One option is to allow AI to reinforce its forces with the Mark-V units available from the respective fabricator(!). Advanced factories simply churn out normal units.
Special Projects include:
- Build Avenger. When complete, send at player homeworld.
- Restore existing Golem wreck for the AI. (Easier way for players to obtain golems, instead of taking the system, if they would rather fight it instead).
- Contact and reinforce human marauders and AI-allied Enclaves.
- Design and replicate Mark-V Dyson Gatlings.
- Build Beachhead device.
- Additional AI-only fabricator+train station to further boost its local reinforcement rate.
- Additional resource train station to further boost its local reinforcement rate.


Player Progress

Observation
The Train network begins invisible(!). Three methods exist for players to expose the network routes
- Maintaining scouting vision of the system tells you a bit about all journeys that enter or leave the system.
- Hack a train and deposit a program (free, but you have to sneak your Hacker to the train). Gain vision of the train itself and its complete journey.
- Hack a station, which tells you about all trains and their destinations involving that station.

This way, players can intuit without scouting the likely locations of rich resource worlds (to take) and fabricator world (to hack or take).

Disruption
Both trains and stations can be raided.
- Destroying a standard train reduces the ability for the fabricator to produce units that generally pressure the player (free threat, special forces, CPA, whatever).
- Destroying the cargo train pauses the progress of the relevant project.
- Destroying the station pauses its contribution. Resource worlds send no metals. AI fabricators can only draw resources from its current system, etc.
- Destroying a significant portion of the network (or continual, effective disruption) will prove that humans are a credible threat, and the AI should respond by switching to reinforcing massively from its intergalactic sources. This is a mini-no return point of game progress.

Note that the 90% standard traffic isn't meant to be completely disruptable. It should be okay for some journeys in this category to be only 1-jump wide - the resource world can only contribute 1 world's worth of resources regardless of distance,

Exploit
If players also end up having stations, then the network can be adopted. Otherwise, several options exist:
- Hack a train and turn it into a Trojan Horse. Your docked fleets disembark at the target station and bypasses outlying defenses.
- Hack a train/station and steal some resources. Alternative income source, where you could sit a cloaked Hacker on a popular route and profit.

[kasnavada]

My feeling is that trains belong in the "ideas for maybe later" part.

That said, I really like the ideas there.

Me -> conflicted.

[Captain Jack]

kasnavada has the right of it. We should be thinking about mechanics that need to be in the game and that already work in AIWC. Chris and Keith and the forum spent weeks if not months at a time trying to make trains fun or even fair and never really managed it. I don't want to see the "with {arbitrary length of time} I can fix this!" pattern start with single mechanics. It was bad enough with SBR.

Astro Trains are one of several mechanics that need rebuilding from the ground up. Save them for when the game is safe and about to come out.

[PokerChen]

On the other hand, as veterans/fans of classic, we are not neutral to the choice  whether AI War 2 should be just an engine upgrade+ streamlining or a deeper reimagining of the core game.  I haven't yet read Chris's inclinations are, so it is worth putting ideas out there to go in the ideas now and ideas maybe later buckets.

Sequels are tricky on the whole to hit home on this respect of how much to change. Everyone who were willing to try classic has probably already bought it, so what would draw on more players? Bour suggestions  bring into AI War the ability to conduct real guerrilla warfare, one of the key missing aspects of classic since 95% of  AI logistics occurred outside the player's galaxy. This talk of possible player actions are fundamental to the design doc, and much more difficult to pull off as an expansion.

So the critical question behind this AI train thread, what we want to answer now as opposed to later, is:

to what degree are players allowed to affect AI production?

[Draco18s]

Astro Trains are one of several mechanics that need rebuilding from the ground up. Save them for when the game is safe and about to come out.

And that's perfectly fine.

I'm just going to let Chris decide if the thread belongs here or there, and what he plans on doing.  Its fine if he thinks the thread is just fine here, but later on down the line says, "Eh, maybe trains won't be 1.0" and saves them for later.