LongLeftFlank Posted July 4, 2017 Share Posted July 4, 2017 On 6/24/2017 at 3:01 PM, Apocal said: An off-map direct fire option would likely be panned by players used to map-edge runs. People already complain about mines in that area during RT MP matches; I can't fathom the howls of incandescent rage that would emerge from unspottable, unsuppressible AT fire. ....and I would personally savour the tears of rage of edge hugging gamey bastidges. But an "OBAT" unit could be a single fixed point on either flank, hence subject to on map terrain, and also time-bounded, i.e. be presumed to be suppressed after a certain point. Just add this to the fantaCM list.... Back on the main topic, I agree that "overstaffing" the computer forces is the primary way to overcome the zombietruppe factor. But I have also long felt that if the TacAI isn't up to simple things like: "Advance along the ditch, not along the sides of it, unless I've told you to Dash" or "'Sir, we've spotted an enemy unit!' Take cover, return fire, then send a team flanking through the trees to the left", or "If my Sherman is on a road and I click a single waypoint down that road, just follow the dang road to it, not a straight line across country, unless I click a waypoint off the road" ... then give the designers Editor tools to pin out sensible (covered) 'paths' to prompt the AI. It's basically a variant of the waypoint tool that automatically repaths around impassable terrain. It basically preplots waypoints, and is invisible to the players but not to the unit TacAIs. Use of such a feature is optional of course, you can always let your zombies be zombies 0 Quote Link to comment Share on other sites More sharing options...
Apocal Posted July 4, 2017 Share Posted July 4, 2017 6 hours ago, LongLeftFlank said: ....and I would personally savour the tears of rage of edge hugging gamey bastidges. But an "OBAT" unit could be a single fixed point on either flank, hence subject to on map terrain, and also time-bounded, i.e. be presumed to be suppressed after a certain point. Just add this to the fantaCM list.... Back on the main topic, I agree that "overstaffing" the computer forces is the primary way to overcome the zombietruppe factor. But I have also long felt that if the TacAI isn't up to simple things like: "Advance along the ditch, not along the sides of it, unless I've told you to Dash" or "'Sir, we've spotted an enemy unit!' Take cover, return fire, then send a team flanking through the trees to the left", or "If my Sherman is on a road and I click a single waypoint down that road, just follow the dang road to it, not a straight line across country, unless I click a waypoint off the road" ... then give the designers Editor tools to pin out sensible (covered) 'paths' to prompt the AI. It's basically a variant of the waypoint tool that automatically repaths around impassable terrain. It basically preplots waypoints, and is invisible to the players but not to the unit TacAIs. Use of such a feature is optional of course, you can always let your zombies be zombies We should probably start with the baby-steps of teaching the AI to not walk directly into machine gun fire or the LOS of known threats they cannot harm, in general. 0 Quote Link to comment Share on other sites More sharing options...
Bulletpoint Posted July 5, 2017 Share Posted July 5, 2017 12 hours ago, Apocal said: We should probably start with the baby-steps of teaching the AI to not walk directly into machine gun fire or the LOS of known threats they cannot harm, in general. There is no AI in this game, only scripting. If the computer-controlled enemies walk into machinegun fire, that's because the human scenario designer has issued an order for group X to move to position B at time XX:XX. This is why the "enemy AI" feels robotic. It is, in fact, programmed just like a robot. 0 Quote Link to comment Share on other sites More sharing options...
Apocal Posted July 5, 2017 Share Posted July 5, 2017 (edited) 1 hour ago, Bulletpoint said: There is no AI in this game, only scripting. If the computer-controlled enemies walk into machinegun fire, that's because the human scenario designer has issued an order for group X to move to position B at time XX:XX. This is why the "enemy AI" feels robotic. It is, in fact, programmed just like a robot. I'm am aware. I have made scenarios in CMx2 before. That's why I wish the AI had the good sense programmers give even the most brain-dead AI in other genres, such as "do not walk straight into, or stand under, a player's killing power." It would greatly ease my concerns if I knew the AI wouldn't keep mashing itself into defense and instead go to ground and at least attempt to gain fire superiority. Other things as well. A basic awareness of what does or does not constitute a threat in front of them, "Player has tank. I will not drive my five scout cars in front of it" or "player has tank riders just behind the edge of the woods. I will not trickle riflemen in there." Knowledge of what are the long suites in any combined arms relationship; "Heavy ATG spotted. Hold tanks' orders to move while allowing infantry forward." Hell, even just ability to blind-fire/call arty on suspected locations would be nice and I know the game already understands the concept of contacts, since it is displayed in game terms. No reason the AI shouldn't be firing on the soft contacts of things like machine gun teams, ATGs, AT teams, etc. that players certainly would. I know I am not asking for impossible here. Possibly unreasonable, given factors I know nothing (and, as a customer, don't care) about, but nothing that hasn't been done. Other games, even ones with one- or two-man programming efforts like CMx2, manage to provide as much. Of course I know the players will still stomp the AI, but as a scenario designer of a defense scenario, I have to actively imagine ways that players will defend any given terrain and create a plan -- using only time as a trigger, no scripting possible -- to make an attack function and look good. Unsurprisingly, the results vary depending on how closely players stick to the script between the designer's ears. For truly adaptive players, frequently the provided AI plans are nonsense wearing clownshoes by the twentieth turn; they are standing bodily in the attacker's start zone, having murdered everything up to that point, just waiting for reinforcements to come in so they can murder them as well. Edited July 5, 2017 by Apocal 0 Quote Link to comment Share on other sites More sharing options...
Erwin Posted July 5, 2017 Share Posted July 5, 2017 AI is probably THE most difficult ability to program. Millions, probably billions have been spent on creating predictive AI to simulate terrorist actions and other phenomena. Not sure how successful it's been as I have been out of that game for a while now. It's not something one would want BF to get sunk on trying to develop, 0 Quote Link to comment Share on other sites More sharing options...
Apocal Posted July 5, 2017 Share Posted July 5, 2017 2 hours ago, Erwin said: AI is probably THE most difficult ability to program. Millions, probably billions have been spent on creating predictive AI to simulate terrorist actions and other phenomena. Not sure how successful it's been as I have been out of that game for a while now. It's not something one would want BF to get sunk on trying to develop, I'm not asking for true AI, with all the predictive abilities, etc. I'm asking that it knows to react to very simple variables, already acknowledged in the game's code, rather than stick to -- and I'm leaning on this example hard because the AI has done it since Shock Force -- running into an area being bombarded by artillery. The game knows where impact areas are; it has to because they are not created dynamically by a direct player intervention, like throwing down claymores in a first person shooter. It knows the danger area of each blast. It knows when rounds will land. It shouldn't be hard to produce a realistic pause rather than entering an impact area. 0 Quote Link to comment Share on other sites More sharing options...
Bulletpoint Posted July 6, 2017 Share Posted July 6, 2017 (edited) One of the very basic things that could be done would be to make computer controlled infantry stop their current movement leg when they encounter good cover along their path. This would to some extent emulate the way a play moves his troops. Currently, the AI doesn't take cover into account when advancing. This frequently means running past sensible cover to sit out in the open until next movement leg kicks in. Another simple improvement would be to make each movement order during an advance shorter, or give the scenario designer a way to control how long the movement legs will be. Edited July 6, 2017 by Bulletpoint 0 Quote Link to comment Share on other sites More sharing options...
LongLeftFlank Posted July 6, 2017 Share Posted July 6, 2017 Yes, I agree that the behaviour of units ought to change based on their "alert" status. It also seems to me that it isn't as hard to put some of this into practice as we all assume. Sure, AI iz Hard. But It doesn't follow that we are powerless to improve unit behaviour or lessen the zombietrupping. "Dash" should be understood as "go directly there stat, don't stop to return fire or go to ground (unless you get hit). But units under "Advance" orders should definitely modify their doubletime pace based on incoming, preferably diverting to cover and then returning fire (but not simply dumping the order). Ideally, unpinned and unRattled units would detach an Assault team to flank and close as the base of fire (hopefully) suppresses the incoming. The flanking pathing reflex might be autoprogrammed in much the same way as panicked units 'bug out' to cover now. The teams seek cover out of sight, but towafd the flanks vs away from the enemy. Once there, they resume the prior AI order to Advance to the destination square, but at a Hunt or Slow pace, not running. 0 Quote Link to comment Share on other sites More sharing options...
A Canadian Cat Posted July 6, 2017 Share Posted July 6, 2017 3 hours ago, Bulletpoint said: One of the very basic things that could be done would be to make computer controlled infantry stop their current movement leg when they encounter good cover along their path. This would to some extent emulate the way a play moves his troops. Currently, the AI doesn't take cover into account when advancing. This frequently means running past sensible cover to sit out in the open until next movement leg kicks in. This is entirely in the control of the designer. If they place the movement end area in a place with good cover the troops will use it. If they place the end area in the open the troops will sit in the open. 3 hours ago, Bulletpoint said: Another simple improvement would be to make each movement order during an advance shorter, or give the scenario designer a way to control how long the movement legs will be. The various movement types or stances (I do not recall the correct term off the top of my head) already control the type and length of movement orders used by the TacAI. 0 Quote Link to comment Share on other sites More sharing options...
A Canadian Cat Posted July 6, 2017 Share Posted July 6, 2017 2 hours ago, LongLeftFlank said: "Dash" should be understood as "go directly there stat, don't stop to return fire or go to ground (unless you get hit). 2 hours ago, LongLeftFlank said: But units under "Advance" orders should definitely modify their doubletime pace based on incoming, preferably diverting to cover and then returning fire (but not simply dumping the order). Correct they do. 2 hours ago, LongLeftFlank said: Ideally, unpinned and unRattled units would detach an Assault team to flank and close as the base of fire (hopefully) suppresses the incoming. Interesting idea and could be an improvement. Only issue is it would be better if two squads from the same platoon coordinated rather than both of them doing this separately. Good idea. 0 Quote Link to comment Share on other sites More sharing options...
LongLeftFlank Posted July 6, 2017 Share Posted July 6, 2017 Agreed that would be even better, but at present there really isn't any (discernible) TacAI functionality at all for coordination between units (other than shared spotting?) My humble ideas were (loosely) based on TacAI / AI program mechanisms that already exist. 0 Quote Link to comment Share on other sites More sharing options...
Bulletpoint Posted July 6, 2017 Share Posted July 6, 2017 (edited) 2 hours ago, IanL said: 6 hours ago, Bulletpoint said: One of the very basic things that could be done would be to make computer controlled infantry stop their current movement leg when they encounter good cover along their path. This would to some extent emulate the way a play moves his troops. Currently, the AI doesn't take cover into account when advancing. This frequently means running past sensible cover to sit out in the open until next movement leg kicks in. This is entirely in the control of the designer. If they place the movement end area in a place with good cover the troops will use it. If they place the end area in the open the troops will sit in the open. For dense terrain, this involves creating a lot of individual move orders. It would be nice to just order a squad to move through an area, and then have them go from cover to cover. As an option of course. 2 hours ago, IanL said: Quote Another simple improvement would be to make each movement order during an advance shorter, or give the scenario designer a way to control how long the movement legs will be. The various movement types or stances (I do not recall the correct term off the top of my head) already control the type and length of movement orders used by the TacAI. I believe that is not correct. The movement types and stances only control whether all teams move at once or if they take turns. Movement lengths are the same, and always "quick", apart from "dash", which uses "fast" (and I think "assault" move order sometimes uses assault move orders too). The stances seem to influence how likely it is for the individual unit to stop and fire at spotted enemies along the route. Edited July 6, 2017 by Bulletpoint 0 Quote Link to comment Share on other sites More sharing options...
A Canadian Cat Posted July 6, 2017 Share Posted July 6, 2017 28 minutes ago, Bulletpoint said: For dense terrain, this involves creating a lot of individual move orders. It would be nice to just order a squad to move through an area, and then have them go from cover to cover. As an option of course. Oh I see you are talking about intelligent placement of the actual move orders generated by the AI between two scripting order areas. Yeah that would be good 28 minutes ago, Bulletpoint said: I believe that is not correct. The movement types and stances only control whether all teams move at once or if they take turns. Movement lengths are the same, and always "quick", apart from "dash", which uses "fast" (and I think "assault" move order sometimes uses assault move orders too). The stances seem to influence how likely it is for the individual unit to stop and fire at spotted enemies along the route. Hummmm, my impression was that the length did change. I guess I'll have to play some extra attention. 0 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.