14 replies to this topic

[Nesutizale]

Lets get one thing out of the way, yes the Ai isn't the smartest but should it be ?

My point is that at some points there are lots of targets around you. Mechs, Tanks, Turrets and VTOLs. With a smarter Ai, would you be even able to take down that numbers of targets without dying?

Lets theorie craft a little bit. Lets say we take MWO as a testbed and put a Lance of 4 above average players against the amount of targets of the easiest mission.

That would be (what I have enountered with a bit of guesswork as I didn't count them all)
- Thunderbold
- 2-3 Locust
- Archer
- Shadowhawk
- Awesome
- Lets say 10-20 tanks
- Lets say 16 turrets
- 4-8 VTOLs

Next we have 3025 tech only with nearly stock builds as MW5 lets you not costumize everything. The Ai is on the average player level.

Would you be able to get through or even still be fun to play or would you die a lot?
What could be done to let the Ai seam smarter without upping the difficulty to a point where only hardcore players still play.

Something that comes to my mind is better pathfinding. That is still quite bad with mechs running through their own buildings or shooting through them.

Ai should at least twist a little bit to protect its sides. Not to often as it makes the game harder but at least try it a little bit..as well as not standing still to be shot.

Ai should also make a run for it when outnumbered or heavly damaged. Mechs are far to valuable to be reduced to slak.

[Koniving]

Welp... There's a few things I can think of.

Friendly AI will shoot THROUGH you to get to an enemy while using an assault mech because they're too slow to stay in formation. So a condition to hold fire or aim away from yourself may help.

Friendly AI completely ignores player aggression. So a clause to recognize player aggression for what it is may help. A similar clause could also be used to introduce the concept of "the third team" and thus create a lot of mission dynamics by having a third party intervene, not initially recognized as a threat but reacted to accordingly when they do become a threat.

When stripped down to just a single shot AC/20, I found the vehicle AI of the one wheeled vehicle I was trying to hit to be really good at trying to evade my shots. (Everything else was dead). It really reminded me as to why factions stopped using Mech "Rifles." But then it suddenly stopped to line up a shot and I instantly killed it. What made that decision even worse was it stopped (while facing away from me) some 100 meters away in order to adjust its turret the extra 25 degrees it needed to hit me since the turret wasn't able to spin fast enough to keep me in its sights. Had the turret been able to traverse faster, or the vehicle decided to strafe around me rather than zig-zag in figure eights, it could have been shooting and evading at the same time.

A mech was treating a destroyed building as if it wasn't destroyed and was capable of providing cover.
It then got confused when I was standing inside the building, and completely lost interest in me. It only renewed interest in me when I shot it, and immediately lost interest after looking at me. (I had already killed my teammates and was on a defend base mission). This repeated again and again. Finally I stepped partially out of the building and suddenly it was time to try and butcher me.

So, for that last one perhaps the ability to recognize a target standing inside a destroyed building could help.

The light medium AI seems to suffer the most. So many Cicadas not taking advantage of what they are.

The biggest thing is there doesn't seem to be much decision making going on. They barely seem aware of each other (it's common to see them shooting through each other). Each is acting independently. What the AI really appears to need is a director or HQ AI.

One that coordinates their decisions. Like telling these enemies to attack the objective while bulkier enemies focus on the player. Such an AI would make bigger missions very impressive.

Another thing is the locational trigger spawns and conditional trigger spawns. When there's no enemies in a direction just a second ago and suddenly they spawn in within 600 meters of you after looking away and looking back as the voice queue alerting you to them doesn't play as well as spawning a long distance away and then approaching you with the voice queue warning you as they enter a certain range.

Edited by Koniving, 27 November 2019 - 03:28 PM.

[Nesutizale]

Agree to all of the above. Nice noticeing.
Also the director Ai is an interesting concept. It kinda mirrors the ability of the player to command his unit.

What I would like to see is that tanks would move more in a formation instead of beeing randomly anywhere. While the single tank is not that dangerouse, a group of tanks firing at the same target can get you in trouble, I think.

Also about the spawning. I wish they would have a bit fewer enemys. Sometimes it feels like endless waves and that isn't good design. At least not to me. A hard fight against overwhelming forces? Okay but it has to end at some point.

[Koniving]

Admittedly the game's AI gets more aggressive when there's fewer of them. Not sure if this is an intentional restriction to empower the player or if more processor time finally gets to the AI when there's not so many of them consuming it.

This post I've recently delivered has a lot more on the director AI concept (in very interesting videos). [Don't worry you won't have to reach much]
It also has a lead-in to a fantastic and cheap-to-buy low budget game that has made masterfully amazing AI. With a spoiler to keep from spoiling why it's amazing if you want to try it yourself the same way I did, blind without knowing anything about it other than it looked cool and was cheap.

[Koniving]

And yes, director (macro) AI does exactly that.
With the exception that while the Director AI would know where you are and what you're doing at all times, it does not share that information with the individual AI. It instead directs patrols and the like near you and lets the individuals 'discover' you on their own.

In the case of MW5, it should have similar behavior to "HQ" in Metal Gear. Directing increased patrols and search parties if allied units fail to respond to checks (if you managed to kill them without them getting a chance to communicate their situation). Once discovered, it should direct forces accordingly while keeping its priorities in check. It would also need to direct the individual AI in their movement patterns to coordinate their attack by creating flanking situations and the like according to force composition.

The AI clearly has a clause or behavior to allow you to retreat. But right now if you trigger this you can retreat at 16 kph and the AI will mostly miss you and allow you to escape so long as you do not look at it or return fire or head in its direction. Basically ignore it while moving away from it and you may trigger it as well. Happens more on easier settings than harder ones.

[Koniving]

Also to an earlier statement I made..
They are, in fact, not aware of each other as entities or allies at all. Each AI actor is self-centered and recognizes other actors only as obstacles.
Given that in this video, (time skipped to the important bit) PGI tells us they made the AI recognize obstacles as something that can be 'shot' or 'walked through' to get to you under a low percentage, this is why they will shoot through each other to get to you. That's not a Locust in front of my Atlas, it's a walking building that's in my way. BOOM!

Shortly after that bug's description and chosen fix, they get into another bug where the level would put you right in the middle of the base and everything activates at once.
This I think is the source of the AI's behavior that the more AI there are in the field after you, the less often they shoot and less accurate they choose to be.

Edited by Koniving, 28 November 2019 - 03:51 PM.

[Koniving]

Watched another good one.
RTS Halo Wars 2 and it's RTS AI that can beat humans and emulate human RTS behaviors without cheating.


Some of the "micro" behaviors here, if exhibited by the AI units under a commander/director AI, would be a huge improvement to MW5's AI by producing some MWO-like human player results.

Time skip to the Macro/Micro strategy/tactics section of the video

[Dimento Graven]

I was surprised at how well PGI modeled the AI to match that of my typical MWO quick play experience.

My lance mates were potaytos, and the enemy was potaytos...

All in all... Very starchy! But fun... I was having fun!

[Hawk_eye]

Nesutizale, on 27 November 2019 - 08:23 AM, said:

Lets get one thing out of the way, yes the Ai isn't the smartest but should it be ?

My point is that at some points there are lots of targets around you. Mechs, Tanks, Turrets and VTOLs. With a smarter Ai, would you be even able to take down that numbers of targets without dying?

Lets theorie craft a little bit. Lets say we take MWO as a testbed and put a Lance of 4 above average players against the amount of targets of the easiest mission.

That would be (what I have enountered with a bit of guesswork as I didn't count them all)
- Thunderbold
- 2-3 Locust
- Archer
- Shadowhawk
- Awesome
- Lets say 10-20 tanks
- Lets say 16 turrets
- 4-8 VTOLs

Next we have 3025 tech only with nearly stock builds as MW5 lets you not costumize everything. The Ai is on the average player level.

Would you be able to get through or even still be fun to play or would you die a lot?
What could be done to let the Ai seam smarter without upping the difficulty to a point where only hardcore players still play.

Is 4 weeks considered a necro?
Anyway, here it goes.

The above (better AI would mean the player can't take on the plethora of enemies) is a fallacy, IMO.

The game only needs to spawn as many enemies, because the AI is so bad.
With better AI, there would be no need for more than, say, a lance of mechs, one of tanks and perhaps a VTOL or two to pose a credible challenge.

It's, again IMO, a kind of cat-chases-it's-own-tail situation
Because of the bad AI, we need lots of enemies to pose a threat and because of the many enemies, we can't make the AI better.


@Koniving good points all around

[Nesutizale]

Koniving, on 28 November 2019 - 04:42 PM, said:

Watched another good one.
RTS Halo Wars 2 and it's RTS AI that can beat humans and emulate human RTS behaviors without cheating.


Some of the "micro" behaviors here, if exhibited by the AI units under a commander/director AI, would be a huge improvement to MW5's AI by producing some MWO-like human player results.

Time skip to the Macro/Micro strategy/tactics section of the video

I wonder if this kind of thing would be possible with Battlemechs. They come in quite a variaty and most of the standart units have multipurpose builds.

Lets take for example the Thunderbold.
LRM15 and Large Laser can be used for long range engagements and if you take the LRM15 alone also indirect bombardment.

3 Med Lasers and an SRM2, while not the most midrange firepower are still good enough to let you fight in midrange combat. Add the large laser into the mix every now and then and you can hurt people.

The 2MG and SRM2 give you options to exploid opened up zones to do critical damage / destroy components. You can also use the MGs very effective against buildings and tanks. They also offer some low DPS if you are close to overheating and the enemy is close by or use it against tanks.

RTS units have mostly 1 kind of attack and maybe a special power, so 2 different attacks and with that also only 2 real options of what to do with them.
The Thunderbold has at least 3 options that themself could be devided into two options again to a total of 6 options.

So one Battlemech has the tactical capabilties of 3-6 normal RTS units. Thats for some simple RTS games the equivalent of half the unit types you can build or close to all the units you can build. I think Halo wars was pretty limited.

When you have a game with about, haven't counted MW5s numbers but lets say 20 different Mechs thats a lot of units that you would need to cover and what the computer has to calculate.
You might be able to reduce that by have more general rules for each mech but I wonder if you could realy have an AI that uses that mix of weapons as effective as a human with just simple behavior pattern.

[Koniving]

Nesutizale, on 26 December 2019 - 03:50 AM, said:

I wonder if this kind of thing would be possible with Battlemechs. They come in quite a variaty and most of the standart units have multipurpose builds.

Lets take for example the Thunderbold.
LRM15 and Large Laser can be used for long range engagements and if you take the LRM15 alone also indirect bombardment.

3 Med Lasers and an SRM2, while not the most midrange firepower are still good enough to let you fight in midrange combat. Add the large laser into the mix every now and then and you can hurt people.

The 2MG and SRM2 give you options to exploid opened up zones to do critical damage / destroy components. You can also use the MGs very effective against buildings and tanks. They also offer some low DPS if you are close to overheating and the enemy is close by or use it against tanks.

RTS units have mostly 1 kind of attack and maybe a special power, so 2 different attacks and with that also only 2 real options of what to do with them.
The Thunderbold has at least 3 options that themself could be devided into two options again to a total of 6 options.

So one Battlemech has the tactical capabilties of 3-6 normal RTS units. Thats for some simple RTS games the equivalent of half the unit types you can build or close to all the units you can build. I think Halo wars was pretty limited.

When you have a game with about, haven't counted MW5s numbers but lets say 20 different Mechs thats a lot of units that you would need to cover and what the computer has to calculate.
You might be able to reduce that by have more general rules for each mech but I wonder if you could realy have an AI that uses that mix of weapons as effective as a human with just simple behavior pattern.

The specifics of RTS games are less important than the behavior.
First the macro behavior of the team as a group and five basic strategies.
"Rush", where the attacking enemy force attempts to get a large force together as quickly as for an all-or-nothing push against a defensive position, systematically destroying all outlying resources, extended bases, and defensive lines.
"Boom", where an enemy force captures/consumes/controls multiple but also the most defendable outlying bases and key structures, preparing for a war of attrition maintaining patrols to keep alert as to what its opponent (the player) is doing and calibrating forces to drain the opponent's resources. (Can work as offensive or defensive)
Map Control, a more aggressive form of Boom that focuses on controlling all important locations and reacts with excessive aggression on any incursions of those locations. (Can work as offensive or offensive-defense)
Fast Tech, doesn't really apply here but seeks to upgrade everything as quickly as possible. This could be done as a mission setting where instead of having massive forces, the enemy force would have a much smaller force of better equipment, better pilots, and stronger teamwork cohesiveness. (Can go either way, but implies smaller but well-defended bases on the defense side).
And Turtle, a defensive strategy that's literally what MW5 already does... lots of low-end units with low-skill to throw at you in waves.

More interesting, though, is the micro tactics.
Pull back: Monitors a small squad and cycles them out when they get damaged to get healed.

Kiting: Keeps their enemies at the edge of their weapons range while either using superior weapons range (Picture support units doing this) or sending multiple units in quickly to attack at ideal range and immediately pull back while another wave rushes in to do the same, allowing attackers to pull out of the player's weapon range to give the player minimal time to respond. (Picture lights doing this, with the coordination that they do it in sequence or pairs).

Push advantage: When the enemies see that they have the advantage, they will begin to push more aggressively. In MW5 you'll notice that mediums-and-up enemies do a lot of posturing where they are less aggressive and hanging back. In MW5 this is the AI using cover tactics, though inexplicably sometimes this happens without any feasible cover. Push advantage would have the AI recognize an advantage over a target and then push to exercise it.

Auto-Garrison: Doesn't really apply here in the traditional sense. But lighter units with longer range capabilities could be recognized by a governing AI as something that should be instructed to hang near larger units, with an inclination for the individual lights to want to take cover behind the heavier unit when it is under heavy attack. A sort of armor sharing.

Scatter: When area of effect attacks or strong offensive attacks are coming their way, the enemy should scatter to produce many targets for you to have to deal with. Ideal for lighter vehicles. Much more noticeable once J. Edgar hovercraft actually get to move faster than 30 kph. (Their canonical cruising speed is over 100 kph).

Group up: Enemies group up against targets to have all their fire coming from one direction.

Focus fire: Focus on one target to the exclusion of all else based on a fitting priority of the situation.

Hold Ground: Something is happening that forces the necessity to hold one's ground even in the face of impending doom. One voice file I heard doing a demolition mission is a perfect example of this: Enemy pilot says "We must stop him now! Right now!!! He is killing civilians!"

Heal: Units actively decide that the time is now to withdraw and heal.

Active Abilities: Likely not applicable in MW5... but one idea is to actively take advantage of unique traits such as indirect fire.

-------------

An overarching AI manipulating the behaviors of the team wouldn't take full control over them or replace their existing AI, it'd set priorities and influences.

For example, in this Alien Isolation clip the Alien AI is rendered visually blind, so it uses audio and touch.

The director AI is omniscient but cannot tell the Alien AI where you are, instead it simply says "Maybe you should search over here...(wink wink)" and when it gauges that the tension is high enough and is supposed to ease off, "Huh, wonder if they went that way" sending the alien to search away from the player. In this particular case, though the director AI keeps redirecting it away from the player (due to the "Tension" of being out in the open), the player keeps drawing the Alien back by making sound. So the individuals make the decisions, the director is only giving suggestions.
For more on how the director AI works..
Time skipped to where it is explained.

As an example in MW5, if Push Advantage is suggested, some (enhanced) individual AI still might decide something else is more important, like "I'm too hurt to push" or "What with? I don't have any guns; screw you I'm getting out of here." A commanding AI isn't going to have any effect on how the individual AI uses its weapons.

It simply gives suggestions, kind of like saying any of the following:
"I think you three should group up and attack this guy. Arrange yourselves like this so you can't shoot each other; then nobody will ever know you're too stupid not to and think that any time you do it is an "accident" of pilot error, and you'll look a LOT smarter even if you aren't."
"Look, I know you're not doing anything so maybe you should go patrol over here. What? No reason, not like there's a player there or anything; I couldn't tell you where the player is but I think you'll find something interesting if you look hard enough."
"You know, I know you're too stupid to focus on just one enemy when you're being hit by five different machines in rapid succession, but I think you'll have a better chance if you focus this guy and just ignore the urge to target the last guy that hit you, okay? You'll look smarter that way."
"Hey, considering you got nothing but LRMs left, maybe you should stop staring and pretending to shoot with the arm that you lost...and get the **** out of here."
"Hol' up there buckaroo, you know running away might not have to be so bad. The player's got a repair bay nearby and it works. Maybe you could pay it a visit while the player's busy...and then destroy the generator after you're fixed. They'll actually think you have a brain and talk about what a **** you are for months to come."

Edited by Koniving, 26 December 2019 - 01:26 PM.

[Nesutizale]

Have you made up your mind what strategic decision (Macro) could work within Mechwarrior?
Because the last examples you have given are to me all tactical decisions (micro) and not an overarching strategie that is given out by a commanding AI.

- Should I fire at another target or groupfire? Thats a tactical decision in a given situation.
- Should I get out of the firing line of my teammates? Tactical decision.
- There is no enemy, should I stay or move on? Mostly tactical. Where to go is more the strategical decision.
- Lost my weapon, should use another one. Again that is a tactical decision.
- Should I use the players resources to my advantage? Also a more tactical decision of the situation.

To me stategic decision of a commanding AI would be for example

- Player pushes me back at grid A9 so I order some units to reinforce
- Patrol lost at grid B5, B4, B3...asume player moves to B2 next so send a force there to intercept
- Encountered player at C4 by light units. Fall back to C5 where reinforcemnts are waiting to lay a trap
- Player has lots of brawlers, move in a units of scouts for target lock and place LRM carriers behind a hill / cover for firesupport

Those, I think, would be more strategic orders an commanding AI could give.
Problem I see and what I mean with my last posting, what units would you use for what if you have so many mixed roll units?
I mean a commanding AI has to look at what it has in units and then select a strategie around those or else it will be stupid again....like trying send LRM carriers into a brawling situation.

Lets take the last example, that is pretty straight forward.
Scout unit: Harraser light tank or Locust
Indirect fire support: LRM Carrier or Catapult, Archer

But what if those units are not avaible.Lets go back to my example of the Thunderbold. A firesupport strategie can still work with a Thunderbold, also only half as good as with a Catapult or Archer. Where do you draw the line of what units are still good for a given strategie when they are so mixed in their design?

[Nightbird]

The reason the game throws so many enemies at you is because the AI is bad. Normally, beating a lance of mechs or two plus some tanks and helis is good enough.

It's not supposed to be

https://youtu.be/fgq0ecMHfzc?t=54

Yet that's all the AI is.

[Koniving]

Nesutizale, on 26 December 2019 - 03:45 PM, said:

Have you made up your mind what strategic decision (Macro) could work within Mechwarrior?
Because the last examples you have given are to me all tactical decisions (micro) and not an overarching strategie that is given out by a commanding AI.

- Should I fire at another target or groupfire? Thats a tactical decision in a given situation.
- Should I get out of the firing line of my teammates? Tactical decision.
- There is no enemy, should I stay or move on? Mostly tactical. Where to go is more the strategical decision.
- Lost my weapon, should use another one. Again that is a tactical decision.
- Should I use the players resources to my advantage? Also a more tactical decision of the situation.

To me stategic decision of a commanding AI would be for example

- Player pushes me back at grid A9 so I order some units to reinforce
- Patrol lost at grid B5, B4, B3...asume player moves to B2 next so send a force there to intercept
- Encountered player at C4 by light units. Fall back to C5 where reinforcemnts are waiting to lay a trap
- Player has lots of brawlers, move in a units of scouts for target lock and place LRM carriers behind a hill / cover for firesupport

Those, I think, would be more strategic orders an commanding AI could give.
Problem I see and what I mean with my last posting, what units would you use for what if you have so many mixed roll units?
I mean a commanding AI has to look at what it has in units and then select a strategie around those or else it will be stupid again....like trying send LRM carriers into a brawling situation.

Lets take the last example, that is pretty straight forward.
Scout unit: Harraser light tank or Locust
Indirect fire support: LRM Carrier or Catapult, Archer

But what if those units are not avaible.Lets go back to my example of the Thunderbold. A firesupport strategie can still work with a Thunderbold, also only half as good as with a Catapult or Archer. Where do you draw the line of what units are still good for a given strategie when they are so mixed in their design?

No intelligent enemy has a singular mindless strategy.
The current AI's insistance on picking a strategy and sticking to it is one of the reasons it seems to be so stupid. Overall, it isn't as stupid as we make it out to be but at the same time, it has many barriers preventing it from intelligently re-evaluating and changing its strategy.

So no, I haven't picked a singular overarching strategy, as the strategy taken by the enemy should be as follows:

• Dynamic to the mission and enemy at hand.
  
• Able to change based on the situation (how are forces? Condition of the objective? Reinforcements coming? What is the value of the objective, is it worth their lives or is it better to pull out? Can lines be drawn? Is the line faltering? Etc.)
  
• Able to change based on its opponent (you'd react differently to a raiding force of lights compared to 4 Atlases going stompy stompy, right? The AI should be able to react differently as well instead of having the same response regardless of what you have) and the condition of the opponent (a healthy attacking force in a defense mission should prioritize defense. A single enemy with no weapons hobbling on a single leg should warrant a hunting party).

It doesn't make sense for an enemy to mindlessly throw everything at the nearest target when it's supposed to be defending an area, completely ignoring the player as the player attacks the key structures.
It doesn't make sense for raiders to throw everything they have at you, rather than taking the **** they came for and getting the **** out.
It doesn't make sense for an enemy that's trying to destroy your base to divert all of its forces to engage you on the wrong side of the map..

But such would be the case if a singular overarching strategy is chosen.

The strategies and tactics I listed were the ones demonstrated in the video as identified tricks that players used, explaining how they worked, and AI is given these as options in order to effect human-like behavior. Then in the video the AI in Halo Wars 2 combines those dynamically based on the situation, the units, and the chosen personality of the enemy commander. One commander is keen on playing long-games, another is brash, aggressive and prone to irrational decisions...etc..etc..etc.

So too, should the enemies we face be fitting of the personalities they are known for. Capellan forces, for example, tend to send in undertrained male forces by significant numbers especially for defense. While I doubt we can revive fallen enemies, it isn't uncommon for a Capellan battlemech that's already fallen in battle to get back up again..and again..and again with a new pilot each time in a single battle. Not unlike what happens in MW5 right now.

Meanwhile, Federated Suns are cocky, they like to egg people on. When surrounded by the Dracs, they had multiple exchanges between the commanders in which the Dracs demanded their surrender and the FS commander's response was along the lines of "We would be happy to accept your surrender, but we don't have enough room or food to imprison and feed such a large force." "No, I meant your surrender." "Oh, no thank you, our position is perfectly tenable." Etc. And naturally they had traps.

Lyran leadership is stereotypically bought, or so the rumors are for why Lyran forces are so incompetent and reliant on brute force... as such regardless of individual ability there's probably a commander nearby that's completely incompetent and too scared to fight himself, or so bold and cocky in the biggest baddest machine there yet completely incompetent in the pilot seat.

Marik is pretty level-headed, prefers to avoid fights and surmise diplomatic solutions, but has well-trained if inexperienced soldiers backed by insane tech, since of all the lost-tech, Marik has the biggest stockpile next to Comstar.

Back to Capellans, you might notice I said the males tend to be thrown out in bulk. For some reason Romano Liao only trusted women with exceptional training, and Liao's elites are female assassins with many devious tricks. That base you're defending has turrets? Huh... your turrets suddenly stopped working, or now have TURNED on you! Suddenly, your lancemate gives a blood-curdling scream, and now a mech of your own design has turned its reticles on you, piloted by a Capellan elite. Jamming? Ryana can't get through to you, and you can't get through to the dropship anymore? Seems like you won't be leaving so soon.

These aren't so much examples of strategies, but personalities, which could influence what strategies and tactics are used. Ultimately the question isn't what overarching strategy every single force ever made should follow... but what forces should have what options available to them, and how those options can reflect what they are.

That's a critical step in the difference from thinking your opponent is just a dumb AI... to immersing yourself in the illusion that you're fighting people.

Some players in the MW5 forums already believe that strange little traits about their lancemates are personalities being shown through rather than flukes of circumstance based on loadout or what-not... Some examples include the affinity to do better for one type of mech over another, or preference for a strategy over another one, or tendency to have a deathwish with specific people. Now notice I said people and not AIs characters. Players have associated these illusions with human traits.

In one of the earliest videos found under my consolidated mod thread, there's a video about making AI simple.. One of the things is that people will fill in the blanks and create their own narrative. If the AI implies a narrative, we'll think it's true and fill in the blanks. Where PGI's AI failed is it did something stupid, which led to finding more stupid things, leading to a never-ending bias of the AI is stupid.

Deserved or not, it has that, because most of the early AI has no personality. It has no human element. Sure you get the occasional distorted bark saying "Surrender to us", etc.. but could surrendering work? Nope, no option and no reaction if you try. Come into a situation where it'd be smart if they ran away, nope they're gonna keep trying to shoot you despite the fact that they have no guns left. What's that, ran into a wall? Lets keep going into the wall.

Its the fact that the AI isn't dropping tactics that aren't working fast enough.. that each AI chooses like one strategy and sticks with it or when they do have a few under their belt they aren't changing to another one when they need to.. that makes the AI seem stupid.

So ideally, a commander AI should not be given a singular overarching strategy. Instead, it should be dynamic based on mission, situation, the opposing force and the player's force; which is further governed by assigned personalities or personas, depicting factions as fans have personified them.

Edited by Koniving, 05 January 2020 - 10:31 PM.

[MW Waldorf Statler]

for the old PS One its gave a mechgame "Carnage Heat" and in this you must self programming the "AI" or better behavior of the Mech ...simple Behavior was lot of orders an logical Links for each operation ,and tahts only by one Maingun.
And whe a order in conflict of Priority of another Order you have a Big problem, and with every object, every enemy that fires at you, the problem increases enormously with the priorities of the program lines.

Edited by MW Waldorf Statler, 05 January 2020 - 10:35 PM.