DUST Search - Tweaking Swarm Launcher parameters

Stefan Stahl wrote:

Kaeru Nayiri wrote:

I'm also giving you 48 hours to necro your own thread / remake it or else I will do it for you, because that is the best swarm launcher operation model I've seen so far suggested. I missed it the first time around.

Thanks for the kind words. I don't think I'm going to revive it though. I came up with an idea. I detailed how it should work. I validated the effect. I published in in the right place. I'm sure the devs read it. Either they implement it or they don't. *shrugs*

The Swarmlauncher has been working like this ever since this game became public. Personally I assume it works this way because in the first implementation of Swarmlaunchers the constructor for swarm missiles was of the type
"SLMissile SLMissile(Vector3d StartPosition, Entity* PointerToTarget)"
instead of
"SLMissile SLMissile(Vector3d StartPosition, Vector3d Direction, Entity* PointerToTarget)"
(there's probably an overwrite where the PointerToTarget is a 3d vector that was originally used for dumb-fire mode).

There is a limited chance that it'd actually be a very complicated change, but I find it much more believable that it's about a day's work from checking out the sourcecode to checking it back in after debugging, but nobody's bothered to do it in the past 3 years.
(Yes, I have a slightly negative outlook - but in my defense I'm not being nonconstructive with my feedback, hence the above suggestion that only requires simple parameter changes.)

[Edit]
I will necro this thread after Fanfest though so Rattati may be able to take a look at it.


Author	Thread Statistics \| Show CCP posts - 1 post(s)
Stefan Stahl Seituoda Taskforce Command 1044	Posted - 2015.03.16 20:19:00 - [1] - Quote Hello, I'm here to suggest some tweaks to how Swarm Launchers work to improve the interaction with dropships, mostly. I'm sorry that this is going to get kind of technical, but we'll be discussing Swarm Launcher stats and their numeric effect on gameplay. The mission statement for this thread is: 1. Make it at least a theoretical possibility to dodge a swarm of missiles. 2. Reduce the amount of time missiles need to catch up to a DS. The fact that Swarms can only ever be outranged, but never dodged through pilot skill, has been a reason for annoyance for quite a while. It'd be great if at least ADS with ABs could do something else about missiles. A similar reason why Swarms can be annoying is because it takes forever to lose them. It takes over 8 seconds of flying away from the Swarmer before the last missile pops and you can go back to work. Count eight Mississippis in your head now - it's an annoyingly long stretch of time to fly straight towards the horizon instead of playing the game. I'll try to stick to the following limitations: - SLs should not be nerfed or buffed against DS or HAVs. - No new features, only parameter tweaks to existing mechanics. Here's the method: If we want Swarm Missiles to catch up with DS more quickly we need to increase their maximum speed and acceleration. In order to give DS a chance to survive we must also decrease their travel range. I've mocked up some calculations in Matlab and found the following data set suitable: - Missile Acceleration 15 m/s-¦ - Missile top speed 85 m/s - Missile travel range 250 m - Missile travel duration 5.1 seconds This may appear woefully fast, but the limited range gives the pilot the same amount of time to react as he currently has. You may remember Rattati's swarm missile calculations for hotfix delta from this thread. I took a look at them and fixed some issues (e.g. measuring the actual time between two missile launches in youtube videos of recent PCs). Here is a graph of the current situation with 400 meter range, here is a graph with the proposed situation with 250 meter range. In both graphs you can see that an ADS with an AB has to eat at least two volleys. If the pilot hits full acceleration within about 1.5 seconds he'll be out of range before the third volley hits. This is the same with the new numbers as before. But all of this now happens within 5.1 seconds of the last volley launch, so this goal is achieved. In the graphs above I used the following parameters: - DS acceleration 12 m/s-¦(with AB) - DS top speed 50 m/s - Current stock SL acceleration 12 m/s-¦ - Current stock SL top speed 60 m/s - Duration between two SL missile launches 2.3 seconds (I looked at several youtube videos of people using prototype Swarm Launchers in PCs. Nobody managed to launch 3 volleys in under 7 seconds.) With those numbers set out we can look at maneuverability or missile turn rate. For this next bit I will interpret "SL missile turn rate" as degrees/second. I haven't seen any alternative explanation that made sense. This used to be at 90, but was reduced to 70 with hotfix delta, without much noticeable effect. I analyzed the issue in Matlab and noticed that it is necessary to reduce turn rate to about 50 -¦/s in order to make it possible to evade a SL missile in an ADS by trying to orbit the missile. Here's a graph of what currently happens. We're looking top down on a Swarmer in the center of the picture and a DS being 125 meter north of him. Blue is the missile's trajectory, red the DS's trajectory. This is the situation as it generally happens with a Swarmer on top of a building. I avoided a 3D scenario to make visualization easier, but in 3D the math works out the same. What you see is the initially static DS trying to orbit the incoming missile counter-clockwise. Due to the high turn rate the missile just impacts the DS after 4 seconds. Here is what happens when you use a turn rate of 50 -¦/s. You may notice there not being any impact. That is because the missile doesn't hit the DS. The DS is taking a weird looking trajectory because it is programmed to maximize angular velocity. After the missile wizzes past coming from the swarmer (about 100 meters in front and 50 meters left from the swarmer's point of view) the DS slows down so the missile misses again on the way back. Afterwards the missile tries to complete the turn but runs out of range before reaching the DS again. The morale of the story is that with a turn rate of 50 -¦/s a current-day ADS with an AB has means to avoid being hit by a Missile other than flying away. The above graph has been achieved with the following parameters: - DS acceleration 12 m/s-¦ (with AB) - DS top speed 50 m/s - SL acceleration 12 m/s-¦ - SL top speed 60 m/s - SL turn rate 50 -¦/s - Impact occurs when the distance between missile and DS is lower than 10 meters (The graph looks very similar if you use the SL speed stats suggested in the paragraph above. I posted it this way so both changes can be discussed individually.) (To be continued in the next post. Wow, hasn't this become a wall of text?)
Stefan Stahl Seituoda Taskforce Command 1044	Posted - 2015.03.16 20:19:00 - [2] - Quote As a net effect of these changes I expect: - Some qualified ADS pilots will be able to evade SL missiles on purpose. - Some SL operators may once in a while miss a shot in certain situations (e.g. DS passing overhead at full speed). - I expect that the time between launching a missile and the +75 WP for doing damage will decrease significantly. - I don't expect any change in the SL's ability to kill shield DS. - Armor DS may have to eat 3 volleys in a slightly smaller timeframe than before (7 s vs. 10 s), if they decide to flee too late, and thus have less time to repair between the volleys. - I don't expect any change at all towards the SL's capability to kill HAVs as those will be too slow to notice the change. - I have no idea whether LAVs will be able to dodge missiles now. I wouldn't even know how to look at making an informed estimate here. I guess the only ones that can take a look at that are the devs on their test rigs. After that wall of text I assume there will be questions. If you'd like to comment on the goals, ask questions about my method or which way to go with parameters, feel free to ask. Please stay on topic, though. P.S.: You actually read that?
Stefan Stahl Seituoda Taskforce Command 1045	Posted - 2015.03.16 20:42:00 - [3] - Quote Thanks for actually taking the time to read. I agree with your conclusion that evading missiles isn't something you can do all day. At best, after an impact, you can set yourself up to dodge the third volley, hopefully with a shield booster ready so you can then face the Swarmer at full health. This limited usefulness is actually why I think it's a good idea. If the new standard way of dealing with SLs would be to dodge them continuously we'd have very unhappy SL users and a new FOTM. This seems like something very skillfull pilots could get an extra kick out. It also looks like something that will make SLs less thoughtless as you'll actually have to think about whether your missiles are going to make it to their target in this situation.
Stefan Stahl Seituoda Taskforce Command 1048	Posted - 2015.03.16 21:21:00 - [4] - Quote Nothing Certain wrote: What is the flight ceiling? I know I often can't lock on because DS are too high, that is 175 meters, if all the DS has to do is go 75-100 meters to outrange a swarm they are not going to be shot down. Again, thanks for reading the wall of text. I specifically designed the acceleration and top speed numbers so that DS won't be able to escape any easier. This shouldn't change at all if I did my math right. (Any validation would be appreciated!) I recon HAVs will have an even harder time outrunning Swarms. The missiles will be faster after the change than before. That should actually help you. The trade-off is that in some situations you'll have to hold off launching a volley on a passing LAV or DS until the angular velocity is low enough for your swarms to get them. Kallas Hallytyr wrote: I applaud your efforts, and I love your workings out especially considering the lack of definitive information about the subject. Would it be possible to look at further narrowing the turn angle? My thinking is that dodging a Swarm volley needs to be reliable enough that it's a reasonable option for the pilot to choose, otherwise, as Alena said, the better option is always to retreat out of range of the lock-on and return again once fully repaired. So, what would 40 d/s look like? What about a higher initial velocity too? From the SDE over at Stuff514 Swarms have a 5m/s initial velocity. A higher initial velocity might space the volleys out enough that a pilot has a greater time to react to follow up volleys, making evasion a more likely course of action. Regardless, I always appreciate the time and effort you put in Stefan, so thank you for doing this :) First of all, thanks for the kind words. The lower the turn rate the better the chances for the DS. Qualitatively nothing changes, you just get more leeway as a pilot and the Swarmer ragequits faster. Personally, as a NDS pilot I'd love to also have the option to dodge, but I'm trying to be cautious here. I don't want a unreasonably large change to cause HAVs to become unhittable on uneven terrain or AB'd Pythons being able to dodge three consecutive volleys of missiles. My philosophy is taking many quick babysteps in the right direction rather than sweeping changes. (Such as buffing armor hardeners by 60% while simultaneously nerfing shield hardeners. ) [Edit] Here's the plot for 40 -¦/s. Also, I don't have any mathematical prove that maximizing angular velocity is the optimal way of dodging a missile. Through experimentation I found maneuvers that can in special cases evade missiles at higher missile turn rates, but I couldn't identify a generic rule to implement. That is another reason to be conservative with the turn rate. People will quickly identify optimal maneuvers that may be better than what I'm looking at.
Stefan Stahl Seituoda Taskforce Command 1052	Posted - 2015.03.16 21:47:00 - [5] - Quote Kallas Hallytyr wrote: [quote=Stefan Stahl]What about the initial velocity? How would a change to initial velocity affect lock/refire rate and reaction windows? My thought is that it would space them a little more, giving pilots a bigger reaction window between volleys. Oh, and quick question: how easy is it for you to knock out different simulations? Do you just need to modify a few numbers and bam, or do you need to rebuild each one? If the former, great, easier; if not, then I won't pester you about doing more simulations It's actually super easy. The longest part is saving the file, uploading it and writing a forum post about it. Here's the distance graph for the initial condition 50 m/s SL speed. In the other type of graphs above the DS starts hovering stationary and then attempts to orbit the missile as soon as it is launched. That's also the reason why the missile always impacts the DS when launched at high initial speeds. The travel time gets very small and the DS has no time to build up any angular velocity. It'd be difficult to ask a DS pilot to maintain angular momentum towards SL launchers he doesn't even know exist yet, so this isn't something that's easy to implement. It'd be awesome in conjunction with directional launching though (shameless plug), as it'd make hitting near targets difficult and hitting far away targets comparatively easy - very good behavior for a Caldari AV weapon.
Stefan Stahl Seituoda Taskforce Command 1055	Posted - 2015.03.17 09:20:00 - [6] - Quote Juno Tristan wrote: I am concerned that there could be 3 sets of swarms in the air before I know they're incoming, but if swarmers have to face the target when firing it could increase the time between the 1st and 2nd volleys. Could you model 2 swarmers in the new versus current setup? I wouldn't want that ttk to increase With a single Swarmer those suggested numbers shouldn't change anything about how much damage you have to tank if you decide to flee. Two swarmers (right next to each other) firing at a hovering DS would also behave the same under the new and current speed parameters. You will eat at least 4 volleys of Swarms if you AB away after the first impact. So no change there. During one of the posted catmaDB dumps it was revealed that Swarm Launchers appear to have a parameter that dictates how far away you can look from a target after lock on before said lock is lost. It's currently at 90-¦. On top of that there was a parameter that appears to dictate how quickly the lock is lost after the first criteria is struck, which is at "1.0" if I remember correctly - I interpreted it as seconds when I read it. Fumbling with those parameters would surely help make SLs more logical to operate. I'm sorry but I don't have a link to that source right now. Maybe somebody else knows what I'm referring to. Vulpes Dolosus wrote: One thing that might be a problem is how swarms actually track their target. Your models, I assume, show the swarms consistently tracking where the target is at the moment. However, and I'm not sure on the specifics, swarms seem to have a "tick" between changing directions. I've also heard it described that they will go to where ever your ship was and follow that path so long as they're able. (If someone else knows a bit more, I'd appreciate the additional input) Again, I'm not sure exactly how missiles operate, but it might change how we approach this problem. Your ideas are really good, and I appreciate the effort. Hopefully we can find some balance. You're right, I assume accurate tracking at a rate of 20 ticks / second. From my limited experience with coding netcode I assume that many of the oddities of how Swarms appear to behave are down to discrepancies between the serverside simulation and the clientside simulation. This is however purely speculative. I don't think we have any means to actually figure out how Swarms behave in detail. Any additional knowledge would be very helpful though. Breakin Stuff wrote: Instead of shortening the range, alter the turn radius to 20 degrees per second, forcingwide-angle turns that arelikely to result with terrain collisions and running out of flight time before they can rebound, while still qualifying as a "guided" missile. The reduced range is mostly there to decrease the time between launching a missile and seeing whether the pilot or the swarmer won. The speed and range stats aren't immediately connected to the turn rate change. They interact a bit when looking at the DS-dodging-a-missile-graphs, but in a small way. I'm hesitant to reduce turn rate too far because I don't want to make SLs useless against LAVs. Maybe a SL should have trouble tracking a LAV going at full speed over rough terrain, but I didn't want to dive into that topic. If a dev finds some time to look at this on a test rig I'd definitely recommend looking at very low turn rates too. Alena Ventrallis wrote: But here's the thing: If I try to dodge the volley, I have a decent chance at making it, but also a decent chance at getting hit. And even if I dodge the swarm, by the time I have stopped my momentum and repositioned, the swarmer has reloaded and the process begins again. Or I can simply fly out of range and be clear of the threat of the swarmer. Running is still by far the best option. Honestly, the only way I see a way to handle swarms that doesn't involve running from them is to add countermeasures like flares or lock-on jammers. Other than that, running will always be a better option to preserve your dropship. I'd like to see what happens with this change. Maybe nothing changes - in that case we only lost the effort of changing a few parameters. Maybe pilots figure out a way to make dodging worthwhile - in that case we gained more interesting gameplay. However I'm not opposed to the having countermeasures and stuff like that at all. The above suggestion is just a minimum-effort way of improving gameplay. That doesn't mean you can't do other things additionally.
Stefan Stahl Seituoda Taskforce Command 1063	Posted - 2015.03.18 00:05:00 - [7] - Quote Nothing Certain wrote: In my opinion, swarms versus ADS is not that imbalanced. I agree there. That's why evasion of missiles shouldn't be something that happens often enough to change the balance between AV and ADS. Just something that rewards excellent skill or punishes launching swarms at really inopportune moments. If this turns out to not have any negative side effects - e.g. on hitting LAVs with swarms - one could then further reduce turn rate, but that'd be something for a second step.
Stefan Stahl Seituoda Taskforce Command 1069	Posted - 2015.03.18 13:12:00 - [8] - Quote Kaeru Nayiri wrote: I'm also giving you 48 hours to necro your own thread / remake it or else I will do it for you, because that is the best swarm launcher operation model I've seen so far suggested. I missed it the first time around. Thanks for the kind words. I don't think I'm going to revive it though. I came up with an idea. I detailed how it should work. I validated the effect. I published in in the right place. I'm sure the devs read it. Either they implement it or they don't. shrugs The Swarmlauncher has been working like this ever since this game became public. Personally I assume it works this way because in the first implementation of Swarmlaunchers the constructor for swarm missiles was of the type "SLMissile SLMissile(Vector3d StartPosition, Entity* PointerToTarget)" instead of "SLMissile SLMissile(Vector3d StartPosition, Vector3d Direction, Entity* PointerToTarget)" (there's probably an overwrite where the PointerToTarget is a 3d vector that was originally used for dumb-fire mode). There is a limited chance that it'd actually be a very complicated change, but I find it much more believable that it's about a day's work from checking out the sourcecode to checking it back in after debugging, but nobody's bothered to do it in the past 3 years. (Yes, I have a slightly negative outlook - but in my defense I'm not being nonconstructive with my feedback, hence the above suggestion that only requires simple parameter changes.) [Edit] I will necro this thread though after Fanfest so maybe Rattati may be able to take a look at it.
Stefan Stahl Seituoda Taskforce Command 1094	Posted - 2015.03.30 08:12:00 - [9] - Quote Stefan Stahl wrote: Kaeru Nayiri wrote: I'm also giving you 48 hours to necro your own thread / remake it or else I will do it for you, because that is the best swarm launcher operation model I've seen so far suggested. I missed it the first time around. Thanks for the kind words. I don't think I'm going to revive it though. I came up with an idea. I detailed how it should work. I validated the effect. I published in in the right place. I'm sure the devs read it. Either they implement it or they don't. shrugs The Swarmlauncher has been working like this ever since this game became public. Personally I assume it works this way because in the first implementation of Swarmlaunchers the constructor for swarm missiles was of the type "SLMissile SLMissile(Vector3d StartPosition, Entity* PointerToTarget)" instead of "SLMissile SLMissile(Vector3d StartPosition, Vector3d Direction, Entity* PointerToTarget)" (there's probably an overwrite where the PointerToTarget is a 3d vector that was originally used for dumb-fire mode). There is a limited chance that it'd actually be a very complicated change, but I find it much more believable that it's about a day's work from checking out the sourcecode to checking it back in after debugging, but nobody's bothered to do it in the past 3 years. (Yes, I have a slightly negative outlook - but in my defense I'm not being nonconstructive with my feedback, hence the above suggestion that only requires simple parameter changes.) [Edit] I will necro this thread after Fanfest though so Rattati may be able to take a look at it. This is the promised post-fanfest bump from vacation. Please don't expect a fast response during the next two weeks.
Stefan Stahl Seituoda Taskforce Command 1104	Posted - 2015.04.12 12:04:00 - [10] - Quote Echo 1991 wrote: Can you explain why you think swarms need to be changed? The only time ADSs die is if they sit there too long. As Kaeru said it's really about being Swarms vs. vehicles being more fun than "be unkillable if you hit the AB quick enough or be doomed if you hit it too late". I'd like there to be more options and this suggestion aims to introduce some fringe cases where you might be able to dodge missiles. The above suggestion only allows dodging in a limited amount of scenarios. You never have to dodge, you can also just run away after the impact as we currently do. As such I don't think this will change the Swarms vs ADS balance at all. The aim is rather to see whether we can make dodging missiles a thing. If that works out we can then - in a second stage - think about things like increasing SL range while further reducing SL maneuverability so ADS can be threatened by SLs at a great range but are unlikely to be killed by them. That's actually how Forgeguns work and it's much better gameplay than SLs are right now. A Forgegun could potentially kill you at up to 300 meters if you're a stupid pilot, but if you keep moving it will barely scratch you. With a SL right now no amount of pilot skill will keep you alive as long as you are within lock-on range. That's what I'd like to change. Kaeru Nayiri wrote: Also Stefan, check my first post where I asked a specific question, I'd be very interested in hearing your answer. Kaeru Nayiri wrote: Imagine the world of Dust514 as it was intended. This includes VTOL jets as well as bikes and bipedal exoskeletons, all of which were planned at the beginning. If we had all of that, do you think it would make sense for the dropship to be able to dodge the swarms meant to catch up to, and destroy, the jets? I originally skipped this question because it's extremely speculative. But since you seem to have an interest, here are my two cents: 1. A current day ADS with the AB on is already pretty nippy. At an acceleration of 12 m/s-¦ it takes an ADS 2.3 sec to accelerate from 0-100 km/h in any horizontal direction. Any vehicle that accelerates much faster than that will really stretch the player's suspension of disbelief as it'd look - quite frankly - ridiculous. Thus I don't think there'd be weapons that are effective against ADS with AB but ineffective against jets. 2. The less the number of categories of weapons the better. What I mean by that is that it wouldn't be enjoyable to have anti-scout guns, anti-medium guns and anti-heavy guns on our dropsuits with each being ineffective towards the other types. In a similar vein it wouldn't be fun to have anti HAV weapons, anti DS weapons and anti jet weapons. While it could be made to make sense from a lore perspective I really think gameplay trumps lore in such questions. On the general topic though, I'd like to make some further suggestions on Swarm parameters, but I can't validate whether my model is correct, so I don't know whether the effort would be worth it. Most importantly I still don't know whether swarm missile maneuverability is actually measured in -¦/sec. I assume so, but at this point I can't continue until I know. If anyone had any pointers towards that, that'd be great.

Stefan Stahl Seituoda Taskforce Command 1140	Posted - 2015.04.30 17:31:00 - [11] - Quote bogeyman m wrote: Stefan Stahl wrote: With a SL right now no amount of pilot skill will keep you alive as long as you are within lock-on range. [...]As a solo swarmer, I can reliably get two volleys in the air before a DS inevitably hits it's hardeners (after the first volley hits) and flys away, regens and returns.[...] With "as long as you are within lock-on range" I meant to imply "... and stay there". As in: Either you hit the magic AB-button early enough and always survive or don't and always lose. I don't think swarms are in any way too powerful against DS or ADS right now. They are just boring gameplay. Hence the above suggestion. Also, thanks to Mr. Apocalyptic Destroyer for the bump.
Stefan Stahl Seituoda Taskforce Command 1175	Posted - 2015.05.12 11:07:00 - [12] - Quote CCP Rattati wrote: Stefan Stahl wrote: P.S.: You actually read that? I did, and this is my favorite post of the month. Of which month? March, April, or May? Humor aside - if you'd like more detailed feedback please give us more technical info on how swarm missiles (and the launchers) work. Most importantly (to me): What is swarm missile turn rate? Degrees/second? Radians/frame? Flips/pancake?
Stefan Stahl Seituoda Taskforce Command 1	Posted - 2015.06.20 19:40:00 - [13] - Quote Loyal Glasses wrote: I am sorry but the topic does not interest me. The only reason I am here is due to the fact that I read the title as twerking swarm launcher parameters. Good day. This is my favorite reply so far.
Stefan Stahl Seituoda Taskforce Command 1	Posted - 2015.07.19 07:56:00 - [14] - Quote Thanks for the bump, Apocalyptic Destroyerr. @Adipem: Most likely none. I don't think HAVs need the ability to dodge missiles effectively and their limited speed and acceleration won't allow for that with any useful set of numbers for the parameters in the original post.
Stefan Stahl Seituoda Taskforce Command 1	Posted - 2016.01.17 09:49:00 - [15] - Quote Slayer Deathbringer wrote: bogeyman m wrote: Breakin Stuff wrote: Faster swarm travel means more time between impacts to react oddly, But the faster the swarms, the less degrees-per-second turn radius they should have. at 85 m/s I'd actually recommend dropping the turn radius to 30 degrees/sec so that dodging the swarms is not only possible, but if a rebound shot is made by the warheads they don't get to pull it more than once. The warheads need to be seekers that home in on the target. they do NOT need to be boomerangs that come back multiple times for seconds and thirds. I would really like one of those boomerang swarms... That's Swarm Launcher Operations level 6, right? yep the Lai Dai Boomerang Swarm Launcher That's actually exactly what happens when you do dodge missiles right now. You can dodge the first impact, but the missiles just turn around and hit you right back. That's because hotfix foxfour implemented the turnrate-changes without the changes to missile speed and duration. At least the deployment of 50-¦/s max turnrate showed that we're not having a problem with LAV tracking. If there ever will be another hotfix discussion I'll propose a further reduction to 40-¦/s and a change to missile life-time and speed.