Aces High Bulletin Board
General Forums => Wishlist => Topic started by: Tilt on June 24, 2012, 11:21:26 AM
-
Please...........
It seems to me that AI GV's would make a massive difference to arenas like AvA.
An AI GV war could be set to run in virtual stalemate with missions alternating such that gv's arrive in conflict at a battle field (think like Kursk). Players now interact over and on this battle field to change the balance of the battle.
This may happen in many ways....I suppose it starts with ground attack AC taking out enemy GV assets...then air superiority would be required for attack ac to operate over the battle field............ then fighters are required to achieve this......none of the air borne stuff would be AI it would be players............. however players could also add their weight to the ground war by taking out enemy gv's and disturbing the balance of the battle.
Setting and timing missions would require an element of skill borne of establishing the best criteria in terms of the opposing mission set ups......but all this starts with AI GV's and develops from there.
BY using a ground war as the AI base we can generate "consequences" such that fronts can move and targets can be captured/lost and eventually wars won/lost....players contributing to this victory as they would in the MA.
-
I dont understand your idea :cry
-
I don't quite understand either. Are you asking for the 20-25% of AHers who play in GVs to be replaced by AI?
-
He says yes
-
It would be better to have AIs in FSO.
-
I have yet to see a pilot good enough to counter "drifting off lazily to the left".
-
AI would need a MASSIVE rework for this to be even remotely feasable. Far more than in aircraft, position and terrain are often deciding factors in a GV fight.
They would have to recognize good cover, likely approach routes based purely on visual appearance of terrain from just their position (if they have knowledge of terrain superior to that of even the most seasoned tanker, they would be TOO smart, even if they were so dumb as to reverse out of cover and start driving around in circles when an aircraft came within 6k), think tactically and strategically in an agressive but intelligent manner if they are to represent the Germans (they could just swarm forward like lemmings and fire randomly at the German panzerklein to represent the russians ;)). But above all, they would need to shoot both accurately and imperfectly.
The biggest obstacle is that they would need to THINK, not just perfectly and correctly analyze the situation based on perfectly accurate information, and then coldly and logically formulate a course of action to most effectively counter the enemy's action. They would need to be susceptible to human emotions, over-analyzation, and errors in judgment.
-
AI would need a MASSIVE rework for this to be even remotely feasable. Far more than in aircraft, position and terrain are often deciding factors in a GV fight.
They would have to recognize good cover, likely approach routes based purely on visual appearance of terrain from just their position (if they have knowledge of terrain superior to that of even the most seasoned tanker, they would be TOO smart, even if they were so dumb as to reverse out of cover and start driving around in circles when an aircraft came within 6k), think tactically and strategically in an agressive but intelligent manner if they are to represent the Germans (they could just swarm forward like lemmings and fire randomly at the German panzerklein to represent the russians ;)). But above all, they would need to shoot both accurately and imperfectly.
The biggest obstacle is that they would need to THINK, not just perfectly and correctly analyze the situation based on perfectly accurate information, and then coldly and logically formulate a course of action to most effectively counter the enemy's action. They would need to be susceptible to human emotions, over-analyzation, and errors in judgment.
in other words I Robot
-
Doubtful AI for GV's could be accomplished, we have to many trees, hills, areas you can't get too.
It can be done with aircraft and the AI, however ground vehicles I don't see it happening any time soon. Far to much work involved, especially coding the AI.
AI's run on a given script, whether to attack or defend they cannot react to changes in an environment - rather only be scripted to make changes - i.e lose a platoon it draws replacements from other areas etc.
The work involved, I dont think so not any time soon.
HTC does not have the resources, as it would take away from the next 30 projects just to draw up the plans and ideas.
-
in other words I Robot
Yes, I-robot is what we need at minimum. However, really this will never be a workable idea, if we want the ai tanks to be capable of more than just holding in place to be destroyed by long-range sniper fire by panthers or Tiger II's, or to swarm forward like a horde of moderately pissed off lemmings. At least not in HiTech's life time.
The issue is that untill we create an AI program that spaces out when it's supposed to be covering a flank, experiences mental fatigue, gets bored and starts being perfunctory in its scan for enemy vehicles, gets excited, suddenly becomes suspicious, etc, then it's either going to be far, far, FAR superior to even the best player (because it will be operating with perfect thinking all the time, with perfect information on enemy positions, etc).
That or it's going to be several orders of magnitude more stupid than even a bad-mediocre player (because it's going to be limited in its thinking and vigilance by the program).
Basically, untill we create AI as in actual intelligence, ambushes will have either a 100% success rate, or they will have a 0% success rate.
-
will the ai tanks claim on 200 that the other side are nothing but campers and noobs?
semp
-
I wouldn't mind having a few AI GVs in town just patrolling the streets. Should be a simple code right?
1. Follow the paths/streets.
2. End of street, turn around and repeat.
3. Tank within line of sight gets shot at.
4. Troops within line of sight gets shot at.
-
I wouldn't mind having a few AI GVs in town just patrolling the streets. Should be a simple code right?
1. Follow the paths/streets.
2. End of street, turn around and repeat.
3. Tank within line of sight gets shot at.
4. Troops within line of sight gets shot at.
I helped work on a game while back quite close to Close Combat that never got developed, one of the ideas that was implimented was the map was "Scanned" For movable paths, all walls/buildings/trees were marked with a "indestructible" tag - this way the tanks could target buildings, however were not allowed to drive over them - in a sense if it got to a brick wall (a building) it would simply turn and move around it. In theory it worked fine, the Tanks were able to move on their own, now the problem was trying to add basic rules like keeping the front of a tank towards the enemy - didn't always work (50% of time)
the Ai did have some logic to it, for example it would use a variable path drummed up, it would never take the same path once, orders were also not hard to impliment, meaning 1 tank could be attached to a platoon or it would roam separate.
All said and done it still came up far to short as far as the AI was concerned, much as we poured script after script into it, it would still try to attack a Tiger from the front with an M4(75) by itself, even when it did group up tanks to attack it the Tiger was in "defensive" mode in which it would be sitting some 4-800yards away from everything and simply pick off an entire platoon of tanks that tried to rush it.
Whole goal was to focus strictly on ground warfare in a game thats not even 4000 yards long by 4000 yards, you had basic orders like attack, defend, advance and retreat.
Honestly best option is what Titanic said, and add basic tanks in a town setting (where the roads are free to roam and not further) this would be a major addition, I would say they could work on the "Spawn" principle where every 45 minutes 2 tanks spawn for a max of 6? Depending on the countries ENY would factor in the type of tank it gets, whether T34/85 or M4(76). I have absolutely no idea if this can be done, or scripts added for it I am assuming it can considering there already is an AI built in for aircraft.
I helped work on the scripts for the game but it was pretty much out of my league, I have seen recent advances in AI scripting in the past 2-3 years enough I do believe it can be done, one game I play all the time called War in the Pacific AE has some pretty awesome script designs.
The Ai has a time frame when it can execute scripts based on Time and troops available, if a criteria isn't met then it simply cannot run the script. Basically scripts like "Attack Port Moresby" Date start 420501 Date End 420801. Naval Support would be TF 1 or the KB, Airborne could be used or Rear Guards (japanese elite ground combat units).
Mission type: Surface Combat w/ Assault, Sze would be #2 Large - Response type would be Maximum, Trigger base Port Moresby, forward base is Rubaul.
Basically this tells the AI its ok to use elite units, between 05/42 and 8/42 attempt an assault on Moresby using the KB for carriers and protection and a large surface TF to clear the path.
What does this mean? basically if you know your history between 5/42 and 8/42 the japanese AI will make an attempt on it - same for midway etc.
This being said I can see Surface Task forces being made AI as they simply travel to the nearest base to "bombard" and retreat or engage warships.
If a real player occupies a cruisers 8 inch gunner then its disabled for the AI, however its able to act as a task force and do simple tasks.
-
The issue with a system similar to Titanic's is that a simple LOS detection system would be inherently flawed in Aces High. How far out will it see a tank coming with 100% reliability? Will the detection be directional, determined by the direction the commander is facing? Will detection be a simple RNG that randomly decides if a tank is detected or not, or will it be based on the percentage of the vehicle that is visible to the drone's commander? Will drones in one side of town respond to gunfire on the other?
A lot more has to go in to it, unless we're content with having nothing more than mobile auto-ack that ignores aircraft. Hell, it would be about as effective to place a couple of '88s in town, and program them to fire on any tank that gets within 1500 yds or so.
-
The issue with a system similar to Titanic's is that a simple LOS detection system would be inherently flawed in Aces High. How far out will it see a tank coming with 100% reliability? Will the detection be directional, determined by the direction the commander is facing? Will detection be a simple RNG that randomly decides if a tank is detected or not, or will it be based on the percentage of the vehicle that is visible to the drone's commander? Will drones in one side of town respond to gunfire on the other?
A lot more has to go in to it, unless we're content with having nothing more than mobile auto-ack that ignores aircraft. Hell, it would be about as effective to place a couple of '88s in town, and program them to fire on any tank that gets within 1500 yds or so.
Line of sight isn't that much of a problem, the Ack guns already have an LOS in which they shoot, prime example of this is drive up to a town and hide behind a hedgegrow, then drive out in the LOS range, 1-2 seconds the ack gun starts firing then back up - it stops.
Tanks would work on the same principle, however ack guns have a range of 6k over a base? Ground vehicles would have to be decreased to so many yards and a line of sight. Otherwise they will simply start shooting at tanks spawning in if there's a LOS.
-
Whats so hard about camping them around a spawnpoint? It will be just like the real thing!
-
In aces high, a system like that is flawed (for GV's at least). There shouldn't be 100% odds of detection, especially if we want to do tanks patrolling town, and not just stationary guns.
-
In aces high, a system like that is flawed (for GV's at least). There shouldn't be 100% odds of detection, especially if we want to do tanks patrolling town, and not just stationary guns.
Also you need to model for gunner accuracy. Also I detect enemy gv's within earshot 100% of the time when I turrn my engine off.
-
The above raises the obvious challenges...... Regardless (IMO) however tackling many of those challenges are very possible with not a lot more tech than that required for ac AI. I to see some other challenges daunting and would. Wonder how they may be solved......
Certainly trees hills buildings etc. (even other vehicles) can be avoided by a simple addition that makes a faux perimeter to the model and steers the GV to avoid collision. If this perimeter disc were at an appropriate height it would also cause the vehicle to steer clear of unassailable gradients.
AI Gv' would still have to follow (as close as terrain and obstacles allow) mission set way points and so core routing does not require any level of intelligence on behalf of the model.
Even in the 90's you could set AI ac in AW to various levels of alertness, accuracy and morale and this could be randomised at various levels across any group of mission generated vehicles such that AI did not always detect enemy, shoot enemy or even advance boldly or run away in a constant and repetitive manner.
Equally now AI mission parameters can be set regarding secondary targets (engaging targets of opportunity etc.) to give the AI missions an opportunity to engage other gv'ers as they are encountered.
Even now icons are revealed to us when enemy vehicles are in line of sight (in a field gun and once upon a time in a vehicle) so the AI can be similarly coaded to see or not see enemy and this can also be influenced by the "-alertness" setting.
We hear other gv'ers now and this is really generated by simple data which relates to if the other gv is engine on and it's relative location. We may assume that the AI would be able to identify the engine note of freindly or foe. And so would use the same core data to determine the position of enemy with an accuracy set by the alertness setting.
The big challenge to me is how to use local cover to stalk and ambush a foe. I don't see a solution to this that can be accomplished opportunistically from within the the AI COAD. I can see how upon hearing an enemy gv that is not in line of sight the AI either attacks all guns blazing (high morale setting) or switches engine off and waits for the foe to appear (medium morale setting and firing with respect to alertness and accuracy setting) or just plain runs away (lowest morale setting).
So IMO the above works sufficiently to create an AI driven battlefield it may miss some intelligent use of cover but would produce a conflict over which real players can interact without certainty of the AI reaction vis the randomised (or degree of randomised) alertness, accuracy and morale settings.
The problem for HTC IMHO is not the insurmountable technical challenge but (as in all wishes) the commitment of resources against the perceived return in player fun.
-
Also you need to model for gunner accuracy. Also I detect enemy gv's within earshot 100% of the time when I turrn my engine off.
Like I said, if we want to do tanks patrolling the town (engines on), they shouldn't have 100% detection, even if the tank is sitting in the open, so long as its something like 3000+ yds out.
They will also need to have to dial in the range, only really acquiring the exact range after 5-6 shots. Their accuracy should also decrease with range, but not in a linear manner.
Detection should not be purely Los based, with a randomization of who gets detected based on the "alertness" of the unit in question. When you tell a player that an RNG just decided that the AI drones would detect his panther because his AA mount was poking up above a hill 1500yds from town, and put a couple of rounds in his flank, but not detect the Tiger II trundleing along out in the open an extra 200 yds back, that's when you start getting the whines.
Basically, what is proposed makes them of rather limited value for an AI + player vs. player battle. To be of any real use in a defensive action, they need to have the intelligence to pull back and relocate, they need to have the intelligence to set up ambushes, and use hull-down tactics. If we want them to be REALLY usefull, we'll need them to know when to counter-attack.
To be of use in an offensive action, they need to know when to push a flank through movement, when to force a flank guard to retreat through presence, when to hang back and snipe, and even when to charge.
I would rather HTC either do the job right, and give us drones we could use for more than patrolling and early detection (because they would need to be in a Panther to pose a threat to a Panzer IV), or not waste their time.
-
Woah....too complex...just slap the same AI that is in the field ack onto some tanks, and slave their movements to the streets of a town. I was just thinking that they'll add a slightly extra challenge to base takers, or give friendlies some extra time to defend. :aok
-
Nice....I like that idea.
-
It seems to me that AI GV's would make a massive difference to arenas like AvA.
Any AI might make a massive difference to AvA. I'm hoping that the new AI mission system will allow us to pre-load several missions and counter missions to run throughout the day.
We could create escorted bombing missions with an opposing intercept mission. As the majority of real players would take up the fighter slots, you will always have the risk of running into real fighter pilot enemies.
If a GV AI system could be implemented the same way, players could fight alongside and against a mix of AI and real players.
I know the GV aspect would be much more difficult, but it would be great if it could someday be implemented.