|
Post by ryujin on Sept 24, 2016 22:45:24 GMT
I noticed a couple things with the IR missiles: The sloped armor at the front is useful and all, but the missiles should be blind without a delicate seeker head upfront. The seeker would also likely be easily disabled by lasers or impacts, reducing the kill time by point defense. Missiles seems to behave like early 60's-70's missiles, chasing targets and crashing into the hottest point. Modern IR air to air will lead the target to intercept, and then try to hit near center mass of the whole heat blob, not just the engines. This is especially true for future missiles which would likely use Imaging Infrared seekers like the latest air to air missiles to recognize the target, ignore flares, and if the resolution is good enough hit specific points (it's closer to EO tracking like a maverick than older IR missiles). For example from the ASRAAM (that's the best missile name ever): "When combined with digital signal-processing and imaging software, ASRAAM is able to see individual areas of its target, such as engines, cockpit or wings. The picture is very similar to a monochrome TV picture, and gives the missile excellent long-range target acquisition, even against employed countermeasures such as flares or similar pyrotechnics." www.raf.mod.uk/equipment/asraam.cfmSo the IR missiles should probably have: -A seeker upfront required for terminal guidance (bigger fancier ones could allow component targeting), making them more delicate and heating sensitive (IR sensors need to be cool). -Mid course command guidance followed by terminal IR (kind of like an IR AMRAAM) which means all the missiles could be targeted at specific ships and then the IR takes over when close for fine tuning). Not guaranteed to hit your ship of choice if there are several close targets, but will point it in the right general direction). -Center of heat blob impact option (hottest point impact could remain a useful option for engine killing missiles) -Lead pursuit on moving targets (they don't seem to do that right now? Hard to tell) -Random weave evasive maneuvers options like modern antiship missiles if you want to burn deltaV. Not a good idea against lasers or agile targets, but would make them harder to hit with rail guns. All of these could be cost and behavior trade-offs when creating a missile. You could make a small, cheap, tail chasing missile swarms or a capital killing high deltaV missile with a large, high quality seeker and evasive maneuvers. It would also make lasers very good anti-IR missile weapons while upping the lethality of missiles and increasing the resistance of high end missiles to flares. Not to get too far off topic, but active radar terminal guidance would let you get your pointy missile noses back and laser resistance at increased cost, energy use, and losing specific component targeting (probably just a center mass hit).
|
|
|
Post by Crazy Tom on Sept 25, 2016 1:27:24 GMT
Yeah, the stupidity of the current missile AI is well documented.
|
|
|
Post by bletchleygeek on Sept 25, 2016 7:23:24 GMT
That was a very interesting post ryujin.
I am the moment going through the blog and there's a pretty illustrative discussion on the post "Sensors and Countermeasures" regarding the resolution you can get at the engagement ranges typical in space
With that level of detail the problem of intercepting a moving target at an angle and at a considerable speed may be just intractable for several reasons:
- if the target aspect angle is about 90 degrees, then the relative speed in the sensor field could be easily 200 or 300 pixels per second, along with dramatic changes in the shape of the object as the aliased projection over the discrete pixel lattice changes. Such a setting would challenge known efficient image segmentation and object tracking algorithms (the same algorithms used in the H.264 codecs most VOD services out there use to compress video).
- leaving aside the ability for an automated system to track targets at the typical relative speeds that may arise from a target that actively tries to dodge, there is this thing about estimating trajectories. Errors in the estimation of the target trajectory are to be expected, and they're going to be pretty massive. Again the scale of distances and speed may mean that an early estimation error cannot be corrected on time due to orbital mechanics or reasonable assumptions on software capabilities (like producing meaningful firing solutions in microseconds) and control operating conditions (the hardware latency being such that it updates the actuators fast enough).
This is not to say that the game AI cannot be improved. If Zane "solved" stuff like the above, besides publishing several excellent research papers, he could just rock up at Lockheed Martin HQ with COADE on a laptop and expect being offered a job for life paying better than a US senate posting...
|
|
|
Post by plastic1d3 on Sept 25, 2016 8:11:07 GMT
The biggest problem is that missiles often seem to home in on a point behind the target ship rather than the ship itself, even if the ship has a disabled engine.
Edit, after some further testing, I think this was because I wasn't setting up good intercepts. Missiles still seem to like the back of the ship (probably because that's where the engine and big radiators tend to be), but actually hit. Had a nice total missile kill which destroyed everything but a ring of aft armor when the ship took a devastator nuke up the tailpipe.
|
|
|
Post by ryujin on Sept 25, 2016 17:59:18 GMT
Going with the assumption that the positions of all ships are known, my guess is that the missile would receive command guidance from the ship up to the last 100 km or so before impact. You wouldn't expect the IIR to guide from 1000km unless you have some crazy seeker. Since you can give missiles commands this doesn't seem to be far off from how they work now.
As far as tracking, that's a good point on relative speed. However I think this should be a bit less of an issue for terminal guidance as it should be start roughly tracking the ship and it just needs to hold it(if not you've already failed). At which point the ships ability to change relative velocity is going to be limited for most ships at maybe 2-3Gs acceleration. It seems plausible since the RIM-161 SM3 uses terminal IR to hit ballistic missiles going about ~6km/s. So maybe? I don't know enough to say if that would work and if there'd be enough prediction to help out the IIR with where the target should be.
Lead would be tricky and ultimately you'll always be able to burn out or reduce the missiles deltaV. Long range shots that don't leave the missile a lot of spare deltaV are always going to be sketchy. At long range given the time involved you could probably be pretty conservative with the command guidance and make the other ship really commit to burns to change it's orbit, not just wiggling the ship and having the missile instantly react.
Missiles in terminal guidance would probably not care about orbits too much and aggressively burn at a simple lead on the target using something like proportional navigation, keeping the image steady in the seeker with the assumption it's coming in already set up for an intercept (which it is in COADE). It'd be high deltaV in the terminal phase, but would be relatively simple and fast. It seems other methods would require a lot of thrusters/reaction wheel mass on the missile or breaking lock as it rotates to burn in different directions.
Just my $0.02, tuning the missiles and modeling systems would be quite a bit of work to get right. I'm just hoping for a bit better, not perfect.
@plastic: I've also seemed to have some more luck using fly bys and telling missiles to briefly scatter if the situation allows before homing to try spread the impact points/angles. Seems to be a bit finicky, but there's probably and ideal way to set the current missiles up for success.
|
|
|
Post by bletchleygeek on Sept 26, 2016 13:59:06 GMT
Going with the assumption that the positions of all ships are known, my guess is that the missile would receive command guidance from the ship up to the last 100 km or so before impact. You wouldn't expect the IIR to guide from 1000km unless you have some crazy seeker. Since you can give missiles commands this doesn't seem to be far off from how they work now. As far as tracking, that's a good point on relative speed. However I think this should be a bit less of an issue for terminal guidance as it should be start roughly tracking the ship and it just needs to hold it(if not you've already failed). At which point the ships ability to change relative velocity is going to be limited for most ships at maybe 2-3Gs acceleration. It seems plausible since the RIM-161 SM3 uses terminal IR to hit ballistic missiles going about ~6km/s. So maybe? I don't know enough to say if that would work and if there'd be enough prediction to help out the IIR with where the target should be. Lead would be tricky and ultimately you'll always be able to burn out or reduce the missiles deltaV. Long range shots that don't leave the missile a lot of spare deltaV are always going to be sketchy. At long range given the time involved you could probably be pretty conservative with the command guidance and make the other ship really commit to burns to change it's orbit, not just wiggling the ship and having the missile instantly react. Missiles in terminal guidance would probably not care about orbits too much and aggressively burn at a simple lead on the target using something like proportional navigation, keeping the image steady in the seeker with the assumption it's coming in already set up for an intercept (which it is in COADE). It'd be high deltaV in the terminal phase, but would be relatively simple and fast. It seems other methods would require a lot of thrusters/reaction wheel mass on the missile or breaking lock as it rotates to burn in different directions. Just my $0.02, tuning the missiles and modeling systems would be quite a bit of work to get right. I'm just hoping for a bit better, not perfect. @plastic: I've also seemed to have some more luck using fly bys and telling missiles to briefly scatter if the situation allows before homing to try spread the impact points/angles. Seems to be a bit finicky, but there's probably and ideal way to set the current missiles up for success. I'll need to think about this and try some stuff in the game, ryujin. Note that I haven't said that things cannot be improved, just making a point (in a humorous way) that the control problem conveyed by automated tracking and guidance of ordnance is already quite hard at the speeds and geometries that make sense within Earth atmosphere, let alone in deep space where our experience solving such problems is close to zero. In any case, I am not either sure either to what degree guidance systems are modelled in COADE, there maybe stuff regarding the software/hardware capabilities that is being fudged and it may be just a matter of tweaking some magic numbers here and there. I find interesting that remark you make about the RIM-161 SM3 being capable of intercepting monsters like the TOPOL (sp?) ICBM surprising... if that's correct it's indeed a great thing for the United States, but I would say that the interception is feasible because of the aspect angle of the incoming missile isn't very wide (e.g. like less than 45 degrees when looking from continental US at ICBMs incoming from central Siberia, for instance).
|
|
|
Post by Acatalepsy on Sept 26, 2016 15:24:14 GMT
I think that point defense and maneuvers, rather than decoys, should generally be the defense against missiles. The current situation is that any fleet with flares and even modest laser or defensive gun armament is completely immune to missiles even while engaging enemy capitals; even without flares or modest point defense, most missiles can be avoided with negligible defensive measures and even a little bit of maneuvering. Any movement perpendicular to the missile's starting location essentially guarantees a miss. And rather than high velocity intercepts being the strength of missiles, they tend to be a weakness because high closing velocity makes the missiles more prone to simply shoot past and not have the dV to come back and actually intercept (I've also had situations where swarms of missiles I launched with 1 km/s closing velocity began an interception AROUND a target, guaranteeing that all of them missed).
This could be solved by giving missiles better interception logic (and making sure that they always began an intercept at sufficient distance to use it). A proper pursuit curve would make having sufficient point defense relevant against missile armed enemies, and the use of maneuvers (orbital and tactical) more relevant to defeating missile swarms, by either running them out of dV, or generating favorable closing distances that give PD time to work.
I would honestly just get rid of flares - technically, they're not very plausible (the logic for defeating flares has been around for a while, it's hard to imagine that future space missiles won't use it, among numerous other reasons), and strategically, they make the game poorer rather than richer by eliminating missile strikes as a serious concern. If they are to be valuable for the game, either they need to be limited in effectiveness. If missile logic gets modified, maybe make it so that each flare only soaks a small number of missiles before the rest figure it out, or maybe each flare can only sucker each missile for a limited time, wasting some dV, before they figure it out and get back on course.
Also, since this is my first post, I wanted to thank qswitched for the amazing work. You can tell how much a pack of nerds love something by how willing we are to nitpick it to death.
|
|
|
Post by bletchleygeek on Sept 27, 2016 0:03:47 GMT
Also, since this is my first post, I wanted to thank qswitched for the amazing work. You can tell how much a pack of nerds love something by how willing we are to nitpick it to death. I disagree about some of your points (flares not making sense, interception logic being a trivial matter) but I cannot agree more with this one
|
|
|
Post by ryujin on Sept 27, 2016 0:43:13 GMT
I'll need to think about this and try some stuff in the game, ryujin. Note that I haven't said that things cannot be improved, just making a point (in a humorous way) that the control problem conveyed by automated tracking and guidance of ordnance is already quite hard at the speeds and geometries that make sense within Earth atmosphere, let alone in deep space where our experience solving such problems is close to zero. In any case, I am not either sure either to what degree guidance systems are modelled in COADE, there maybe stuff regarding the software/hardware capabilities that is being fudged and it may be just a matter of tweaking some magic numbers here and there. I find interesting that remark you make about the RIM-161 SM3 being capable of intercepting monsters like the TOPOL (sp?) ICBM surprising... if that's correct it's indeed a great thing for the United States, but I would say that the interception is feasible because of the aspect angle of the incoming missile isn't very wide (e.g. like less than 45 degrees when looking from continental US at ICBMs incoming from central Siberia, for instance). A few of the most modern long range SAMs have ABM capabilities like the S-300/400 and Patriot. However due to the constraints of having to fire a missile up through air the intercept range on ballistic missiles with these SAMs is fairly small (and likely not close to 100%). They're more of ballistic missile point defense, from what I understand the missile needs to be on the terminal phase to hit within a few miles of the SAM, which allows the intercepting missile to make a pretty much head on intercept in time. Until there's space based or hypersonic missiles, these systems are more useful for defending an area from conventional theater ballistic missiles like the Iskander. Not going to be much good in a full nuclear exchange blanketing the country with clusters of MIRVs, but better than nothing. In COADE I think a number of the guidance details can be fudged while being close enough. It's more the math for the intercept it flies and having some sensor vulnerabilities. The control for long range intercepts with orbits and such is going to be the most complex, but right now it's handled by the player plotting burns. The terminal intercept shouldn't be more complex than in atmosphere. There's no drag and gravity will be minor enough that you can ignore it and burn right at the intercept point for the terminal bit. If you're willing to spend a little more deltaV you can simplify things a bit. Also, since this is my first post, I wanted to thank qswitched for the amazing work. You can tell how much a pack of nerds love something by how willing we are to nitpick it to death. I disagree about some of your points (flares not making sense, interception logic being a trivial matter) but I cannot agree more with this one Agreed, haven't nerded out this much in a while.
|
|
|
Post by bletchleygeek on Sept 27, 2016 3:00:48 GMT
In COADE I think a number of the guidance details can be fudged while being close enough. It's more the math for the intercept it flies and having some sensor vulnerabilities. The control for long range intercepts with orbits and such is going to be the most complex, but right now it's handled by the player plotting burns. The terminal intercept shouldn't be more complex than in atmosphere. There's no drag and gravity will be minor enough that you can ignore it and burn right at the intercept point for the terminal bit. If you're willing to spend a little more deltaV you can simplify things a bit. That's a very good point - maybe similarly as one does on LEO, the control can be mapped onto a simple Local Vertical/Local Horizontal frame. That would simplify things enormously. The challenge is that then you would need to come up, with 1) a closed loop controller that guarantees interception provided certain constraints on relative speeds and distances are met and 2) an open loop controller to determine the micro-burns that steer the missile into the region where you can fire the former. I do know from first hand the latter to be hard. Of course, if you have a magical closed loop controller, with easy to meet constraints, the search for the open loop controller may be quite trivial.
|
|
acatalepsy
Junior Member
Not Currently In Space
Posts: 97
|
Post by acatalepsy on Sept 27, 2016 13:19:30 GMT
I disagree about some of your points (flares not making sense, interception logic being a trivial matter) but I cannot agree more with this one I wouldn't call interception logic trivial exactly - especially when you're dealing with uncertain sensors, uncertainty in both own position and enemy vector, and more that the game don't have to deal with - but it can be solved in game, and at the end of the day any interception logic that tried to kill lateral velocity rather than accelerating towards a heat source (again, this guarantees a miss against any maneuvering target) would be a vast improvement on the usability of missiles and make both point defense and orbital maneuvers more vital to defense. I don't think flares can't make sense, but that they're bad for the game and not really logically necessary. The same logic that makes, say, radar-guided missiles not a thing (because the 'metagame' of missiles and countermeasures worked out towards heat seekers) could just as easily apply to flares. Flares as essentially perfect missile protection is boring, and makes missiles barely worth including.
|
|
|
Post by bletchleygeek on Sept 27, 2016 13:36:24 GMT
I disagree about some of your points (flares not making sense, interception logic being a trivial matter) but I cannot agree more with this one I wouldn't call interception logic trivial exactly - especially when you're dealing with uncertain sensors, uncertainty in both own position and enemy vector, and more that the game don't have to deal with - but it can be solved in game, and at the end of the day any interception logic that tried to kill lateral velocity rather than accelerating towards a heat source (again, this guarantees a miss against any maneuvering target) would be a vast improvement on the usability of missiles and make both point defense and orbital maneuvers more vital to defense. I don't think flares can't make sense, but that they're bad for the game and not really logically necessary. The same logic that makes, say, radar-guided missiles not a thing (because the 'metagame' of missiles and countermeasures worked out towards heat seekers) could just as easily apply to flares. Flares as essentially perfect missile protection is boring, and makes missiles barely worth including. Okay, it is true that one thing is "in game" and another "in real life". With such a truth-to-nature sim like this is kind of hard not to confuse them. And you've got a great point re: radar guidance not being cool because of chaff, when flares seem to be as effective at distracting IR guided weapons, as chaff is supposed to be.
|
|
|
Post by Blothorn on Sept 27, 2016 17:34:31 GMT
And I believe that missiles with some degree of spectral seeking/behavioural discrimination (I think spectral seeking got some attention in the blog). I suspect that details are still classified, which tends to exclude tech from the game, but here it seems fairly jarring.
I think that the reasoning about chaff vs. flares is that while they are both very effective, flares are more expensive for any decently hot ship. I see the point of missiles not in actual destruction (although you can achieve it against the unwary) but in creating a threat that forces the enemy to spend mass to counter it. (And I suspect that you can beat the AI by sending 1-2 missiles at a time until they run out of flares. A smarter opponent would trust his point defense against such small waves, so it becomes a question of whether you can overwhelm his point defense with a wave of missiles massing less than the flares he would need.)
|
|
acatalepsy
Junior Member
Not Currently In Space
Posts: 97
|
Post by acatalepsy on Sept 27, 2016 17:42:55 GMT
And I believe that missiles with some degree of spectral seeking/behavioural discrimination (I think spectral seeking got some attention in the blog). I suspect that details are still classified, which tends to exclude tech from the game, but here it seems fairly jarring. I think that the reasoning about chaff vs. flares is that while they are both very effective, flares are more expensive for any decently hot ship. I see the point of missiles not in actual destruction (although you can achieve it against the unwary) but in creating a threat that forces the enemy to spend mass to counter it. (And I suspect that you can beat the AI by sending 1-2 missiles at a time until they run out of flares. A smarter opponent would trust his point defense against such small waves, so it becomes a question of whether you can overwhelm his point defense with a wave of missiles massing less than the flares he would need.) The problem is that, in order for missiles to create a threat that forces the enemy to spend mass to counter it, the missiles need to actually be a threat. Right now, they aren't and can't be. They can't be because they can't hit the target if the target spends even tiny amounts of effort evading, they can't penetrate even small amounts of point defense in limited numbers, and can be suckered by cheap decoys if the target doesn't feel like spending the ~50 m/s necessary to evade. The amount of mass you spend in (ineffectual) missiles is so much more than you could possibly force your opponent to waste, that it's better to just not include them. In order to be a threat, missiles need to be able to zero their lateral velocity, at a minimum.
|
|
|
Post by Blothorn on Sept 27, 2016 19:20:36 GMT
What are you using for point defense? I have found lasers near-useless (multiple optimized 13MW lasers usually getting 1-2 kills before flak missiles enter their danger space), and the only kinetic weapons I have made with an effective range usefully greater than the danger space abused the integrator bug.
I agree that the guidance needs help; I was referring to the missile/decoy balance. (I have had decent results from lining up precise interceptions and reserving a large DV for the terminal phase, but it should not be that hard.) It is not hard to make a 5km/s, 8G missile at much lower mass than a 100MW flare.
|
|