The Ur-Quan Masters Discussion Forum

The Ur-Quan Masters Re-Release => General UQM Discussion => Topic started by: OrzBrain on February 02, 2010, 12:04:08 am

Title: AI and difficulty mod suggestions.
Post by: OrzBrain on February 02, 2010, 12:04:08 am
I was starting to play UQM again, and got to thinking about what would make it much better and more interesting: an AI improvement mod. I should state that I know nothing of programing, so I'm saying sorry in advance if these ideas are impossible to do easily.

(1) The Melee AI could use some basic fixes. The Thradish Torch flame problem is the most obvious. I'm not talking about with just the Thradish, though. The AI doesn't seem to avoid Orz marines, Kor Ah blades, Androsynth bubbles, possibly DOGIs, Earthling missiles, and various other things. Also, the AI has an oddity when you get close to the screen warp around point. It reverses course, and you can use this to hold, say, a Dreadnought across the screen forever while you plink at it risk free with an Earthling cruiser.

(2) AI improvement. An interesting concept from XCOM, which some of you have probably played: Randomness. As I understand it XCOM doesn't have particularly wonderful AI, but it uses a high level of random behavior to fake it. Now in UQM random movement wouldn’t work so well, but if AI modules could be randomly loaded from a small situation specific subset it could make a really nice AI.
For instance, at the moment the AI pursues in most cases. One could make it so if it has pursued for so long without getting in weapon’s range or landing a hit it has a random chance of selecting from a module subselection, say, Pillbox and Grav Warp. In the Pillbox module the AI ship comes to a stop and waits for, say, up to 1.5 minutes for the player ship to come to it. In Grav Warp it breaks off pursuit and heads for the planet, getting a sling shot in a random direction and staying at sling shot speed for a random time while it waits for the player to get in range.

With enough modules and clever randomizing programing it seems to me one could make it much better than it is now.

Also some neat difficulty mods could be added to the main game.

(1) When the flagship speed exceeds the speed of Ur Quan fleets, a random Ur Quan fleet could get superspeed enough to catch up without dragging the flagship down into the mass of others. Not all the time. Just occasionally.

(2) When the flagship weapons get too overpowered if you take it into battle the enemy should spawn with multiple ships in melee -- I'd say from 2 to 10 randomly.

What do you think?

Title: Re: AI and difficulty mod suggestions.
Post by: Death 999 on February 08, 2010, 04:26:45 pm
As for your third suggestion, I think it'd be a lot better if at least an intercept course were attempted, rather than flying straight toward the ship. And if they were similar-avoiding so they'd form more of a cordon than a mass...

Title: Re: AI and difficulty mod suggestions.
Post by: Tiberian on February 08, 2010, 05:33:42 pm
Balancing the game along the progress of the player is evil.

Your suggestion is much like Oblivion, where all enemies level up with the player. That is bad in so many ways... If you think it is good, you are a lost cause.

Another such example is the dreaded "rubber band AI" in most driving games. So the better you drive, the faster your opponents are.

I REALLY hope it is not just me who hates those kinds of things. They take away the reward from being good.

Title: Re: AI and difficulty mod suggestions.
Post by: AngusThermopyle on February 08, 2010, 10:04:13 pm
Agreed about the Oblivion difficulty mechanism. I didn't like how I could basically win the game as a level 1 character.

As for UQM, I do like the idea of  programming the AI to adapt to the player's tactics. For example, let's say the player defeats 10 Dreadnoughts with an Eluder. Is it not conceivable that an  intelligent species like the Ur-Quan would change their tactics? So when the 11th Dreadnought warps in, it pillboxes instead of chasing the Eluder, as OrzBrain mentioned.  If the player defeats the pillboxing Dreadnoughts x number of times, maybe the Ur-Quan try gravity whips/orbiting tactics. If this proves ineffective, the AI could cycle between chasing, pillboxing, and planet shenanigans just to mix things up.

This doesn't really give the AI a decided tactical advantage (as with the racing game example), but it does keep the player guessing as to which AI 'personality' they may be facing in battle.

Title: Re: AI and difficulty mod suggestions.
Post by: OrzBrain on February 08, 2010, 10:15:46 pm
I don't think I'm suggesting a rubber band AI. Also, those options could be made so one could disable them. But it seems to me it makes sense in the game. If the Quans notice you've got a ship powerful enough to blow up a hundred in a row without a scratch, they would try massed attacks. And if they notice they can't ever catch up, one might try redlining his engines occasionally.

I think my suggesting of randomization is much better than some kind of adaptive AI. As far as I've seen those never work, and are hell to try to code. Whereas randomizing gives the impression of intelligence, without the difficulty of creating it. As in XCOM.

Title: Re: AI and difficulty mod suggestions.
Post by: Tiberian on February 09, 2010, 06:17:22 am
I think Angus is on the right track here.