nkenny

@johnnyboy That is weird. My community plays with enhanced movement, and I've seen none of those reports. I will continue to investigate -k

@Aetherblade Must have skipped it in making somehow. Yes. As Duke_SFG says, it is the same for all my steam releases. -k

@Robustcolor That is correct. The *2 is there to create a little deviation, so the units do not spawn perfectly away from the direction of the victim. 5 degrees deviation. 10 - 5 + random (5 * 2) Crates a range of 5-15 -k

A small video demonstrating two new waypoints added to v2.0 and some of the new optional keyboard shortcuts. @Devastator_cm While the AI is more likely to blindfire, nothing has been done to how the AI perceives the world or engages point targets. Perhaps this is a case of false attribution 🙂 also, the modding suggested above may help. -k

Press Escape and select the Camera button to access the new Splendid Camera 🙂 -k

@Ryko My tests have found no great performance costs. I've run countless local tests and two larger scale op tests, 20-30 players-- with a third scheduled for tomorrow. My local tests have found FPS to be _exactly_ the same as vanilla. In fact, I have sometimes seen better FPS with mine. There have been reports of greater memory usage with the mod active. I have not had any consistent evidence to corroborate this; especially not with version 2.0 The lambs danger.fsm overrides (extends and enhances) the vanilla one The vanilla FSM is not run at all. Performance impact, if any, is minimal. Especially compared to the weight of mods like ACE3, RHS and CUP. -k

Here is a follow-up video that further explores the final mature implementation of fleeing and flight in LAMBS danger.fsm version 2.0 Key things to notice: The distinction between fleeing and hiding. Fleeing is a vanilla feature which is triggered when the AI is grossly outmatched and unable to damage its enemy with a weapon. Hiding is a danger.fsm operation which happens both on the individual level (when the AI finds itself without weapons to damage the immediate threat) and the group level (when an infantry leader assesses that there are armoured or air threats present). This video showcases all these matters focusing mainly on responses to Armoured vehicles-- but ending with a small demonstration of helicopter responses. obs: At the time of writing the video is still being processed so resolution may be limited. -k

@slamduck You are correct. Good observation. It is a limitation of how information is shared and gathered by the AI as compared to humans. It is worth mentioning that information was shared to the BTR from the patrol standing next to it. In other tests I've run, without infantry present-- enemy armour is far more inert. One of the big difficulties of a fully universal-- in a terrain and equipment sense-- AI like Arma3 is that we humans come equipped with much more contextual information than is translated to the AI. A human might expect an attack to come from a certain direction. Might pretend not to know. Might have other units hidden or ready to respond. Indeed, in nearly every platoon on platoon engagement style mission the human side is generally briefed with information about expected location of enemy and reinforcements. Comparatively, the AI is dropped in place and expected to respond to best ability. I won't say it can't be done better. But sometimes a level of abstraction is necessary for the AI to appear more human. -k

@MacTheGoon Yes, indeed. The mod works seamlessly on any unit spawned through any means. The only known incompatibility is ASR, which loads its own danger.fsm --- - Some progress, demonstration and discussion on the flight and fleeing mechanics in Arma3:

Not hard // just the player ~ in the direction away from the center _deviation = 40; _pos = _player getpos [_distance,(_pos getDir _player) + _deviation - (random (_deviation * 2))]; -k

So infront of the player? As in closer to center? Not entirely sure what you mean. Try to reverse the _distance though. (i.e., _distance * -1) -k

@LSValmont It has been a really busy couple of weeks has left me a little off the forum. What you describe is implemented in version 2.0. I'll be making a YT video sometime soon. The release itself is soon.tm, moving precisely on schedule. -k

@Devastator_cm While not a priority, we will look into ZEUS support in the future. Specifically via theeasily expanded ZEN mod interface. @Sammael While much of this is governed by cfgWeapons which remains untouched by the mod, there are elements of this in the new leadership manoevure parameters. I am not entirely happy with how suppression works at the moment. Generally it is a good idea, but difficult to automate in a way which is universally applicable-- ie, read by AI. Scripts and modules like the ACE suppression module are very good alternatives. @Antagon Wystrel Ehh. The mod is fully ACRE compatible at the moment, given that AI do not use radios. ;) -k

@Ryko Fantastic feedback. I will look into this a little more closely. There are two other matters at play. 1. At long range, every purely shooting action with LAMBS_danger.fsm enabled comes with a chance of suppression fire. Suppression fire is inherently less accurate than controlled fire. 2. The LAMBS danger.fsm comes with a strong REACTION impulse on first contact. This is why the unit with the mod active attempts to fall back or otherwise seek cover. Had he been in a group, his leader would have mitigated this behaviour by calling for attacks. Additionally, when a shooting event comes up-- forceSpeed is set to 0 at long range. This is complicated by the fact that the command seems to have little effect for some type of movement. In any case, this all merits further investigation.

@PTV-Jobo I find skill settings to be an arcane and byzantine, subject to individual preference and expectations more than any mutually agreed upon formula. Even more so than accuracy settings I find that the weapon plays a major part. I have investigated this in some detail in the LINKED thread. The long and the short of it is that carried weapon play a larger part in simulating certain types of opponent than does any skill setting. Mostly because AI reactions are opaque. LAMBS_danger.fsm mod seeks to address some of that. My own precision and skill settings are in the high 0.8-0.9. This is more a matter of personal machismo than it is any careful experiment. On the nopryl.no server, we're running somewhat lower settings-- adding delays due to client count and internet inertia. I have seen mods which normalise skill settings into more usefully tweaked figured. My own experience with this is limited to reading the code-- I have so far found game play adjusted through enemy equipment to be a a more useful measurement. -k

Panic behaviour is triggered when suppression (from getSuppression) is higher than 0.97 and a random chance (1 in 20 by default) is hit. Because the FSM does not cycle at perfect intervals, the rate which this check is performed will vary somewhat. I consider this fluctuation reasonable. In version 2.0 panic chance is in terms of setting increased, but in actuality is generally lower. As the panic effect is somewhat less invasive and there is generally more time between cycles. Setting lambs_Danger_panic chance very high will indeed remove panic as a game mechanic. I would like to experiment a little more with accuracy related to the constant movement. Offhand I would say that more mobility is a fair trade off for loss of accuracy at close range. -k

@CW4 Carrera Just run the following command line in any relevant init field: lambs_danger_debug_fsm = true; lambs-danger_debug_functions = true; These will be handled with CBA settings in version 2.0 @Ryko Thank you for your extended description. I will attempt to answer in order. (1) and (2), the mod does nothing to vanilla or mod weapon configuration. Hence the perceived inaccuracies or erratic use of weapon capabilities (grenade launchers, burst modes, and so on) are products of other settings. Earlier in this thread and across some of my videos I explore the difference in what is configured and what is not. If you play with @LAMBS_suppression there will in a passive, secondary way be added inaccuracy. This is because AI units loose accuracy on becoming suppressed and this this mod increases the time a unit remains suppressed. LAMBS_Suppression is developed to increase the difference between highly trained troops and lesser ones. It also develops ideas explored in my investigations in AI weapon accuracy (Link given above). If your unit plays with exaggerated, or significantly decreased AI skills (as many do). Your mileage may vary some. This is especially true when CUP configured weapons are used interchangeably with RHS ones. This for reasons explored in the aforementioned link resource. In CQC the movement cycles of the AI of greatly sped up. It may indeed merit further investigation. :) (3) I explore in some detail the design considerations behind these settings in a post a few pages back. Panic is a function of suppression, and suppression is very much affected by skill settings. For lack of consistency of configuration, I have avoided adding more arbitrary links to skill. Perhaps when the mod reaches a higher level of maturity. Like mentioned. Version 2.0 will see many more CBA variables. :) Kind regards, Ken

@CW4 Carrera debug information is unfortunately only shown where the unit is local. Unless zeus is in play, this will most likely be the dedicated server itself. @gerhart I'm unfortunately not familiar with Overthrow. If there is a custom danger.fsm. These may indeed be competing, in the sense that one will supplant the other. Civilians running in place may potentially happen if they are path locked. Once 1.96 is released we will have more ways to avoid this. :) Ken

If you make the mission yourself, I guess you could run a script when a player joins a group. So that if it is greater than a certain size-- you kick him/her out. -k

I think ACE3 has one, configurable via a nifty CBA interface . https://ace3mod.com/wiki/feature/viewdistance.html

// just the player ~ behind wherever he is facing. _pos = _player getpos [_distance, (getDir player) + 180]; // just the player ~ in the direction away from the center _pos = _player getpos [_distance,_pos getDir _player]; In either case you will also need to update the follow line to remove the getPos. // notice the removed 'getpos' private _group = [_pos,EAST,6] call bis_fnc_spawnGroup;

@Robustcolor Code example Function use The nk_fnc_unitSpawn function accepts three arguments 0 : Position of unit spawn 1: Range within which players will trigger script, default is 250 meters (optional) 2: Distance which the enemy group will be spawned, default is 600 meters (optional) On running the function will consider one position. The position may be an ARRAY, a MARKER or OBJECT. It will then check among all players to see if any one of these is within RANGE meters (default 250). If present, it will pick a random one of the players. Behind one randomly selected player, the script will pick one position DISTANCE meters (default 600) away. It will then pick the nearest building and use that as a spawning point. The spawned unit will be given the taskRush mission. example use: ["marker1"] call nk_fnc_unitSpawn ; [ObjectiveTank, 500, 1000] call nk_fnc_unitSpawn; Homework: Firstly, these scripts are written offhand while at work. There is probably an error somewhere. Secondly, why not try to convert the in-file function to one that is preprocessed like fn_taskRush.sqf. :)

Sorry, haven't paid attention to this topic. @RCA3 Those would be pretty sleek upgrades, and I might steal borrow some of those ideas for a future game scenario. For now, I enjoy the simplicity of these scripts. They are also very easily expandable or modifiable to fit needs. As the video mentions, the intention is to hit or reach the players in a consistent manner. @johnnyboy Thank you, that is a great idea! @Robustcolor It certainly is! This is how I run many of my unit spawning scripts. Perhaps the path to the script or some other aspect of the script is wrong? Example: _group = [player getPos [600,random 360],EAST,6] call bis_fnc_spawnGroup; [_group,1000] execVM "fn_taskRush.sqf"; Should work just fine. 🙂

Thanks everyone for the kind words. @froggyluv Regarding subskills. Because this mod does not modify cfgWeapons settings, the actual weapon performance will not be affected as such. Though you will be seeing significantly more aggressive and active AI than vanilla settings. The biggest difference between high and low skill settings, as it related to the danger.fsm, is that I reference suppression variable of the AI. This is MASSIVELY affected by unit skill. In vanilla a 0.2 skill soldiers will be suppressed for 10 seconds. A skill 1 unit, 1 second... In regards to cover, I am hoping to hit a nice medium of cover and mobility. As the video notes, it is unfortunately impossible to alter this variable dynamically (without access to the source code). Perhaps further experimentation will hit gold. For reasons of not wanting to overload the AI, I try to avoid adding additional weapon changes and such to the fsm. Instead letting the default cfgWeapons carry that weight. @rainbow47 The plan ahead looks like this. The 2.0 is feature frozen. We are now only seeking to document and clean up existing features. We're doing a larger scale test of version 2.0 in my community this Tuesday. After which will comes period where we fix any bugs or add tweaks as necessary. Provided there are no massive issues, we're looking at a 1-2 week period. @AirShark In fact, infantry already hide from Air and armoured assets 🙂 I'll make a video about it. As for artillery, I actually haven't thought about adding specific handling of that! @LSValmont and @jandrews We're currently trying to reach a nice compromise. At the current level of access there are unfortunately limits to what we can and cannot do-- though this is massive compared to any other game. So props to Bohemia. I also suspect that CQB is a little too aggressive in the current state. This is really a matter of getting timings and settings right. It is possible to alter the CQB range (that is, at what range the AI consider itself in CQB). Lowering this will increase use of cover in a sense. - nkenny

Thanks Jester 🙂 -- Here is a stream-of-consciousness overview of the tweaks and changes done to the core pathfinding and selection process which will be live in version 2.0