Posted by: Garfunkel
« on: August 12, 2025, 07:52:13 PM »That's a lot of action and good to see the new features working!
Posted by: jahwillprovide
« on: August 10, 2025, 04:57:07 PM »Im looking through your crafts, what is the MK I Heavy Bolter you have on the Aquila-class Lander? I dont think its a railgun because Im not seeing the 1x4 etc. Plasma Carronade or something?
Edit: I figured it out, 1 shot rail lines up with that power
Edit: I figured it out, 1 shot rail lines up with that power
Posted by: Garfunkel
« on: July 17, 2025, 07:43:47 PM »What an exciting start! You definitely got lucky with all those JP ambushes, as you mentioned, things probably would have gone differently in an open space engagement. Eager to see NPR carriers!
Posted by: Ghostly
« on: July 15, 2025, 02:15:41 PM »I think I have figured out a solution (in my head), so when I get time I will code it. We just moved into a new house today so chaos will reign for a few days.
Congratulations on your new home!
Now, I'd hate to derail the comments thread (and suggest things that could potentially delay the update) even further, but it's difficult to keep silent when tactical AI is being discussed. NPRs learning how to scout JPs and properly assault them with squadron jumps would be incredible, yes, but would it work without granting the AI object permanence?
I don't know if it can memorize lost contacts on the strategic level (i.e. if it keeps a list of enemy ships seen in each system recently), but it definitely doesn't remember any lost contacts on the tactical level and is therefore very vulnerable to ambushes of all kinds, and will routinely split its forces to chase contacts whose smaller escorts it recently saw/got attacked by. but then forgot about, or jump into blockaded systems where its ships were already assaulted upon entry, sometimes doing so dozens of times. Not to mention the old peeking problem that was already mentioned here would return if the AI doesn't remember what it sees.
I might be asking for too much here, implementing lost contact awareness does seem like a substantial task, but for combat AI, I'd say this would be the biggest improvement that I could possibly think of.
As for how this would actually look like, I'd suggest the AI should behave like every enemy ship it sees in a system is still there until contact is re-established or a sufficient amount of time (days? weeks? something on the order of double or triple the time it would take for known enemy contacts to approach any NPR colony or fleet, or leave the system, whichever would take longer) elapses. So if it detects 20 hostile ships entering its system, it should consolidate its forces and only create detachments that it deems capable of engaging those 20 ships, and keep its colonies protected while putting in every effort to re-establish contact using its scout ships and possibly scout fighters.
Re-acquiring any contact that was previously seen in a large group of contacts (a fleet) should make it react as if the whole fleet was there until the opposite is proven. If it observes those contacts leave the system via a JP, it should blockade that jump point (as it currently always does), but also consider giving chase right away if it judges its own forces to be superior. There's never a valid reason for a hungry swarm to abandon its pursuit of a slow, tasty survey ship just because it managed to transit a JP while minutes away from getting devoured.
Since I mentioned scout fighters, we've yet to see NPR carriers in your AARs so I don't know if what I'm suggesting here is plausible, but I can say from my own experience that long-range scout fighters are a very powerful detection tool and if the AI could use them to locate enemy forces and maintain contact, it would grant it enough battlefield intelligence to engage the player on far more even terms.
Posted by: Steve Walmsley
« on: July 15, 2025, 10:42:20 AM »Yes, but I am quite sure we had this conversation other times before and after these
https://aurora2.pentarch.org/index.php?topic=8495.msg113856#msg113856
https://aurora2.pentarch.org/index.php?topic=12523.msg157641#msg157641
So I don't really recall why the current solution was applied
The change in the second post was made because NPRs had a habit of jumping through stable JPs and immediately jumping back through, which made it impossible to actually shoot at them. Notably, players cannot do the same - NPRs are able to shoot at a player ship in the same increment after a JP transit.
For my $0.02, if NPRs cannot learn how to scout the JP before transit (which is understandably quite difficult, even for players it often requires odd tricks or breaking roleplay), if a JP is assumed to be hostile they should make a squadron jump, since it isn't as if a squadron jump really costs anything. The trouble, as Steve has noted, is that splitting up NPR fleets for a set of squadron jumps confuses the tactical AI. If Steve can solve that, NPRs will become far more capable at jump points (that, and maintaining a defensive stance over a long period of time when necessary, also not an easy task even for players).
I think I have figured out a solution (in my head), so when I get time I will code it. We just moved into a new house today so chaos will reign for a few days.
Posted by: nuclearslurpee
« on: July 15, 2025, 10:25:27 AM »Yes, but I am quite sure we had this conversation other times before and after these
https://aurora2.pentarch.org/index.php?topic=8495.msg113856#msg113856
https://aurora2.pentarch.org/index.php?topic=12523.msg157641#msg157641
So I don't really recall why the current solution was applied
The change in the second post was made because NPRs had a habit of jumping through stable JPs and immediately jumping back through, which made it impossible to actually shoot at them. Notably, players cannot do the same - NPRs are able to shoot at a player ship in the same increment after a JP transit.
For my $0.02, if NPRs cannot learn how to scout the JP before transit (which is understandably quite difficult, even for players it often requires odd tricks or breaking roleplay), if a JP is assumed to be hostile they should make a squadron jump, since it isn't as if a squadron jump really costs anything. The trouble, as Steve has noted, is that splitting up NPR fleets for a set of squadron jumps confuses the tactical AI. If Steve can solve that, NPRs will become far more capable at jump points (that, and maintaining a defensive stance over a long period of time when necessary, also not an easy task even for players).
Posted by: Froggiest1982
« on: July 14, 2025, 08:10:57 PM »I was reading and waiting to see where this was going. I can see only a couple of solutions that would not end up in a Risk stalemate where you just piling up armies on the other side of a jump point:
- Going back to the old advantage where NPR have no jump shock to deal with, effectively "cheating", or by giving them a reduced jump shock, so that you are still getting the advantage but for limited amount of time compared to the current and player
- AI would send a scout in advance prior jumping and if an enemy fleet is on the other side would decide if the mission they are carrying over justifies the risk and pretty much still ending up wiped out eventually
The problem with 1 is that we are trying to get away from cheating AIs in favor of a more "smart" one. The problem with 2 is that we could go back to the old peeking problem with scouts jumping in and out systems every cycle.
We could have a random zero-time jump shock, valid also for the player.
Yes, but I am quite sure we had this conversation other times before and after these
https://aurora2.pentarch.org/index.php?topic=8495.msg113856#msg113856
https://aurora2.pentarch.org/index.php?topic=12523.msg157641#msg157641
So I don't really recall why the current solution was applied
EDIT: As you will see NPR already worked on a reduced Jump Shock basis.
Posted by: paolot
« on: July 14, 2025, 06:02:44 PM »I was reading and waiting to see where this was going. I can see only a couple of solutions that would not end up in a Risk stalemate where you just piling up armies on the other side of a jump point:
- Going back to the old advantage where NPR have no jump shock to deal with, effectively "cheating", or by giving them a reduced jump shock, so that you are still getting the advantage but for limited amount of time compared to the current and player
- AI would send a scout in advance prior jumping and if an enemy fleet is on the other side would decide if the mission they are carrying over justifies the risk and pretty much still ending up wiped out eventually
The problem with 1 is that we are trying to get away from cheating AIs in favor of a more "smart" one. The problem with 2 is that we could go back to the old peeking problem with scouts jumping in and out systems every cycle.
We could have a random zero-time jump shock, valid also for the player.
Posted by: Froggiest1982
« on: July 14, 2025, 04:45:25 PM »I was reading and waiting to see where this was going. I can see only a couple of solutions that would not end up in a Risk stalemate where you just piling up armies on the other side of a jump point:
The problem with 1 is that we are trying to get away from cheating AIs in favor of a more "smart" one. The problem with 2 is that we could go back to the old peeking problem with scouts jumping in and out systems every cycle.
- Going back to the old advantage where NPR have no jump shock to deal with, effectively "cheating", or by giving them a reduced jump shock, so that you are still getting the advantage but for limited amount of time compared to the current and player
- AI would send a scout in advance prior jumping and if an enemy fleet is on the other side would decide if the mission they are carrying over justifies the risk and pretty much still ending up wiped out eventually
The problem with 1 is that we are trying to get away from cheating AIs in favor of a more "smart" one. The problem with 2 is that we could go back to the old peeking problem with scouts jumping in and out systems every cycle.
Posted by: Steve Walmsley
« on: July 14, 2025, 08:13:12 AM »
- Aurora is very complex so the AI is never going to be competitive with an experienced human player. The AI struggles in many AAA titles with large dev teams and much simpler mechanics - plus I suspect the average Aurora player is smarter than the average gamer, which makes the AI gap wider. All I can really do is keep adding new considerations for the AI and tweaking the code as problems present themselves.
FWIW, I think the AI in Aurora is possibly more competent than the AI in the average Triple-AAA game coming out. Not least since it does, after all, continually improve.
In any case, I think regardless of AI tactical ability, the Vorchan were sunk anyways due to inferior technology and doctrine (or perhaps more accurately lack of superior technology, coupled with a poor doctrine for the situation). But you do like to see when the AI can at least mount a credible JP assault instead of getting walloped by 1/3 their tonnage in defenders. Myself, I'm still searching for that sweet spot, bumping up difficulty can work but can also leave the player in some settings facing faster, longer-ranged opposition perhaps a tad bit sooner than desired.
I agree that JP assaults definitely need improvement. I've been giving it some thought and I will make a few changes. Although we just arrived back on the island and moving into our new house tomorrow, so will be a few days before I get chance.
Posted by: Xkill
« on: July 14, 2025, 02:21:21 AM »General AI shortcomings is why I try to make their digital lives as easy as possible. No kiting, no crazy cheesy designs, refrain from using squadron jumps, etc. Always use stuff that they can understand, like FACs instead of fighters. For me, superior speed is functionally maneuver. All I need it to do is reduce enemy accuracy and increase mine. Either I go to minimum weapons range of the enemy or I retreat. The AI is a lot more formidable under these circumstances.
Posted by: nuclearslurpee
« on: July 13, 2025, 06:54:20 PM »
- Aurora is very complex so the AI is never going to be competitive with an experienced human player. The AI struggles in many AAA titles with large dev teams and much simpler mechanics - plus I suspect the average Aurora player is smarter than the average gamer, which makes the AI gap wider. All I can really do is keep adding new considerations for the AI and tweaking the code as problems present themselves.
FWIW, I think the AI in Aurora is possibly more competent than the AI in the average Triple-AAA game coming out. Not least since it does, after all, continually improve.
In any case, I think regardless of AI tactical ability, the Vorchan were sunk anyways due to inferior technology and doctrine (or perhaps more accurately lack of superior technology, coupled with a poor doctrine for the situation). But you do like to see when the AI can at least mount a credible JP assault instead of getting walloped by 1/3 their tonnage in defenders. Myself, I'm still searching for that sweet spot, bumping up difficulty can work but can also leave the player in some settings facing faster, longer-ranged opposition perhaps a tad bit sooner than desired.
Posted by: Steve Walmsley
« on: July 13, 2025, 05:12:12 AM »Exciting stuff. Looks like the multi-system NPR is working as intended. Forward surveyors are a must here as is a robust auxiliary corps to establish bases as you advance through enemy territory.
Various comments I had from reading this:Probably some others I forgot in reading this majestic epic of an update...
- Will NPRs ever get suspicious that a system holds a dormant JP and resurvey it? It seems like the player could have the option to jump into a system, meet a dangerous NPR, and then just hide behind that dormant JP forever to completely mitigate the threat until they feel like invading, while the NPR doesn't really know how to defend itself in this situation.
- It seems like NPRs in general don't understand how to handle jump points. With the tech + doctrine disadvantage, I don't think this NPR would ever have a chance to win a JP assault anyways, but I don't think they at any point used a squadron jump to get out of beam range and/or reduce the jump shock time to get their weapons online quicker. Either tactic might have enabled the Vorchan to at least do some damage.
- Similarly, the NPR would have been smarter to defend at the JPs where they could fight an engagement better suited to their fleet doctrine. Missile fleets are awful on the attack, but defensively they can mass fire against a fleet dispersed after a squadron jump to at least kill something.
- I notice the Vorchan never combined fleets to achieve superior mass, despite this supposedly being added to the AI some time ago.
- I'm curious if you have other thoughts on NPR performance here. Again, they have a tech disadvantage (or at least a lack of advantage coupled to an ill-fitting doctrine), but between this and the previous Japanese campaign it seems like NPR lack of tactical skill is the major weak spot right now.
- Maybe future campaigns should be played with a difficulty above 100%, so the NPRs generally have more of a tech advantage to counter the player's advantage of not being as dumb as a bag of rocks.
Hoping that this one goes on long enough to meet other NPRs including these mythical NPR carriers we have been promised!
- I don't think there is resurvey code right now, although they can still find the jump point by detecting a transit.
- Jump point assaults are tricky for the NPRs as it stands. I haven't updated their ship designs to take account of the jump point changes, plus NPR fleets have AI that functions differently depending on their role, so splitting them up for a while with independent actions before they recombine needs new AI code. I just need to devote some time to it.
- After losing some fleets, they did position some defensively at jump points. Also, I think something else is going on because fleets keep moving across their Empire into the areas I haven't seen yet. They may be fighting someone else too.
- The NPRs will combine fleets, although I think that is currently a tactical (System AI) choice, not a strategic one (Empire AI).
- Aurora is very complex so the AI is never going to be competitive with an experienced human player. The AI struggles in many AAA titles with large dev teams and much simpler mechanics - plus I suspect the average Aurora player is smarter than the average gamer, which makes the AI gap wider. All I can really do is keep adding new considerations for the AI and tweaking the code as problems present themselves.
Posted by: nuclearslurpee
« on: July 12, 2025, 11:13:37 AM »Exciting stuff. Looks like the multi-system NPR is working as intended. Forward surveyors are a must here as is a robust auxiliary corps to establish bases as you advance through enemy territory.
Various comments I had from reading this:
Hoping that this one goes on long enough to meet other NPRs including these mythical NPR carriers we have been promised!
Various comments I had from reading this:
- Will NPRs ever get suspicious that a system holds a dormant JP and resurvey it? It seems like the player could have the option to jump into a system, meet a dangerous NPR, and then just hide behind that dormant JP forever to completely mitigate the threat until they feel like invading, while the NPR doesn't really know how to defend itself in this situation.
- It seems like NPRs in general don't understand how to handle jump points. With the tech + doctrine disadvantage, I don't think this NPR would ever have a chance to win a JP assault anyways, but I don't think they at any point used a squadron jump to get out of beam range and/or reduce the jump shock time to get their weapons online quicker. Either tactic might have enabled the Vorchan to at least do some damage.
- Similarly, the NPR would have been smarter to defend at the JPs where they could fight an engagement better suited to their fleet doctrine. Missile fleets are awful on the attack, but defensively they can mass fire against a fleet dispersed after a squadron jump to at least kill something.
- I notice the Vorchan never combined fleets to achieve superior mass, despite this supposedly being added to the AI some time ago.
- I'm curious if you have other thoughts on NPR performance here. Again, they have a tech disadvantage (or at least a lack of advantage coupled to an ill-fitting doctrine), but between this and the previous Japanese campaign it seems like NPR lack of tactical skill is the major weak spot right now.
- Maybe future campaigns should be played with a difficulty above 100%, so the NPRs generally have more of a tech advantage to counter the player's advantage of not being as dumb as a bag of rocks.
Hoping that this one goes on long enough to meet other NPRs including these mythical NPR carriers we have been promised!
Posted by: Blacklight
« on: July 11, 2025, 08:42:53 AM »Looks awesome so far, insane to run into a multi system NPR almost immediately. So far looks like youre kicking them around the park, be interesting to see if they manage to turn things around. It definitely seems like just 1 bit of bad timing for you and one of those missile fleets in amongst the hen house could end very badly.
