Some simple AI improvements

For what's not in 'Top Priority Game Design'. Post your ideas, visions, suggestions for the game, rules, modifications, etc.

Moderators: Oberlus, Oberlus

Post Reply
Space Squid
Posts: 50
Joined: Thu May 09, 2013 4:46 am

Some simple AI improvements

#1 Post by Telos » Sat Jan 21, 2017 9:19 pm

I know this game is just alpha, and programming AI is tremendously hard. But here are a few glaring weaknesses in the current AI, together with suggestions that could help make it a bit less bad at this game.

1. The AI often builds way too many troop ships, and not nearly enough military to support them. The simplest fix would be just to shift percentages in the direction of producing military. A more difficult, but better, fix would be to have it estimate which targets it has a realistic chance of invading soon, and build just enough troop ships to take control of those, and then to stop building troop ships until additional targets become realistically invadable. This would cut back on some rare opportunities for an AI to quickly steamroll across multiple planets, but would help a lot in much more common cases where an AI instead loses everything due to lack of military. Much better to establish air superiority and take a long time to control the ground than to lose air superiority and thereby lose everything.

2. The AI is lousy at defending vulnerable ships like troop carriers or small fleets, so experiences tremendous economic drain replacing them, and even in cases where it finally secures a military victory over an enemy planet, it often has lost all nearby troop carriers by the time an invasion becomes possible. A common example of this is cases where an enemy fleet consisting just of scouts and troop carriers arrives at one of your shielded planets, where it can't invade due to shields, and is a sitting duck for the first ship you bring to trigger combat. An ideal realistic solution would probably require assessing the value and safety of all systems, keeping vulnerable ships in safer systems, and repositioning military assets to better protect unsafe valuable systems. A simpler solution, which would still be a great improvement over the current game, would be simply to require that small fleets and troop ships travel only alongside, or at least on the core-world side of, military escorts. It might also be wise to have the AI steer away from fast fragile troop carriers like flux and organic, whose speed it isn't smart enough to make good use of, and whose fragility makes them evaporate quickly. Instead, it'd probably be better served by making much more protected troop carriers that can help to soak damage alongside military ships in a fight and will have a much better chance of surviving long enough to do their job: this means use asteroid carriers more, and even to consider adding a bit of good armor onto troop carriers. The only occasions where I tend to take notable losses versus the AI are occasions where it builds a bunch of asteroid carriers, and stacks them together with planets and/or military fleets, and then the presence of all these heavy sponges keeps the their damage dealers alive long enough for my ships to take losses too. So the AI should basically force more fights like those to happen, while feeding me much fewer free ships.

3. The AI relies too much on scouts (which it controls poorly) and not enough on detection technology, which can automatically see as far as the AI usually needs to see without much opportunity for poor-decision making to mess things up. This is especially true in cases where an AI has already been ambushed by stealthy ships, so should know that it needs better detection technology. After early game, the AI should give very low priority to building scouts, and very high research priority to increased detection ability. Whenever an AI suffers a significant defeat from stealthy ships, this should trigger even higher research priority into further detection technology.

4. The AI responds extremely poorly to stealth. As mentioned above, significant defeats by stealthy enemies should trigger high priority for detection research. It should also trigger high priority for building lighthouses (not a bad idea regardless, just for the speed/unpredictability bonus) and distortion modulators. It should probably also shift priorities towards viewing systems as unsafe, and (hence) towards concentrating ships in larger fleets and/or keeping fleets stacked with planetary defenses, especially vulnerable ships like troop carriers. The presence of stealthy threats should also shift priorities towards building ships with high innate structure and/or heavier armor. Glass cannons are quite useless against stealthy enemies as they blow up before they even get the chance to lock onto a target. In contrast, heavily armored turtles with relatively light armaments can withstand the initial "free" ambush barrage, and then still hit back hard enough to punish stealthy invaders.

Post Reply