Space missiles!

[NFunky]

Hi guys, I have a couple questions about stock space missiles.

I've built a nice little missile with a probe, battery, reaction wheel and four seperatrons.  My first question is, what is the best part to use as a warhead?  I guess I mean which part will cause the most damage on collision with another vessel?

My second, much more important question is, how do I make it guide properly?  The 'hold target' function gets it close, but because everything's moving in orbit of Kerbin, it doesn't fly the exact direction it's pointed.  I've had good luck just manually maneuvering it to bring the prograde vector onto the target, and have had some pretty spectactular collisions, but I'd love to find a way, or a mod, that will automatically allign the prograde to target.  Does this exist?  Can MechJeb2 do this, and if so, what section is it in?  Any other mod suggestions?

[diomedea]

Any autopilot able to be given a simple instruction of the kind of "keep relative speed oriented along the relative position vector" would be able to make what you asked. Autopilots programmed in kOS definitely would make it easily.

However, your target appears "stationary" only when you are very close (Note, this is a very different issue than with moving targets as I described above). The larger the angular distance along the orbit, the larger the error due to how KSP computes relative speed (vector difference from target absolute speed and vessel absolute speed): the farther away, the more the two vectors point in different directions (you certainly observed that making a close R/V with target, relative speed starts high but gets down the more the target is close, even while neither vessel nor target have changed speed). That vectorial difference therefore includes an angular component, that seems an error to be compensated. Quite often players keep trying to compensate that error without reducing relative speed first, ending running circles around the target. Your missile will need to be very maneuverable to cancel that angular error while it approaches the target at high speed. Very possible it may have to use the main engine to provide a lot of vectored thrust, so your autopilot would require a routine to orient the engine thrust accordingly and then to keep firing the engine until angular error is reduced to 0 (which is when target is along the prograde direction, but beware of the apparent error due to distance).

So, I'd not use docking autopilots. They need to keep orientation of the vessel's port parallel to the targeted docking port, and thrust is generally only made out of RCS in translation mode.

[NFunky]

Just an update on some experiments I've been running.  And I'm still having to guide missiles manually, which means missing 80% of the time.

I've tried several parts as warheads, and as far as I can tell, there is very little difference in damage.  However, I discovered that the more parts you can hit at once, the better chance of destroying the target.  So, I gave the 10m inflatable heat shield a try and it made a pretty big difference.

It looks like penetration power is not a big factor in KSP, which makes sense, given the fact that we are dealing with unarmored parts.  So a big, blunt surface is actually ideal.  Not only does the 10m heat shield do more damage, it makes a bit of difference as far as scoring a hit as well.  The large surface area means more like a 50% chance of me manually guiding it to impact.  Only problem is, the 10m heat shield is a big part, even when stowed, so it's not easy to mount on a 0.62m missile.

Still working on guidance.  I've tried several experiments with MechJeb's docking autopilot, but no luck.  I am about to start testing with the manuever planner executing an intercept node.  This is all with an imobile target BTW.

[diomedea]

Warhead: KSP handles explosions (for graphical effect but also damage to nearby parts), but don't expect a part being destroyed on your missile to produce damage to a nearby vessel. While in reality warhead fragments projected all around and the shockwave bring enough energy, KSP doesn't transmit any from explosions to another vessel. KSP however works fine with physical forces, therefore transmission of kinetic energy: impacting another vessel makes an impulsive force F = dV/dT * m, dV/dT is the change in speed with time, m the mass of the impacter. If said force produces a displacement on the impacted part larger than what the physics engine allows, the impacted part detaches from the target (which is generally reason to start an explosion in KSP). So, you have to maximize kinetic energy with the impacter, meaning using a part with a large mass and moving at high relative speed towards target. And the mass of the impacted part is also relevant, if large that part will be able to absorb the kinetic energy without exploding.

 

Guidance: aiming directly at a target position works best with a stationary or slow target. It still can be successful if chasing a high speed target (weapon fired from the target rear). With moving targets at wide aspect angle, kinematic makes for increasing change in angles the lower the distance gets, and invariably the missile won't be able to maneuver fast enough (change its speed vector to keep following target). Which is still acceptable if said missile can explode its warhead when close enough to damage target, but can't work in KSP. Instead, you need to use a guidance system using proportional navigation. Which means, the missile is directed at the predicted target position, movements ratios (missile/target speed components) in each coordinate are the same (proportional), angles don't change. Actually, when a change in aiming angle is detected, the guidance system uses that to further change its own speed vector, until the change in aiming angle with time becomes zero.

Don't know about any add-on for KSP that implements proportional navigation guidance. However, the principle is easy enough so could be programmed in a autopilot using kOS. However please know one thing: the PID based autopilots (used in stock KSP and all add-ons I know about) work fine when goal is to reach and keep a specific orientation, but aren't able to keep pace with a changing value. To make an adaptive guidance system able to compensate movements of a moving target you need a closed-loop feedback system of at least 2nd order (acceleration) (real missiles guidance systems include also 3rd or even 4th order derivatives), on both pitch and yaw axes. Feedback systems require precise calibration to be fast while avoiding overcompensation, their response is tied to timing (or phase) of the error signal used as input.

[NFunky]

This is really awesome info, thank you!  I don't really know kOS at all, but it sounds like that's the only way to go with guidance.  The fact that I can hit the target with manual orientation makes me think it really shouldn't be difficult to program.  I'm firing at "stationary" targets, but because everything is in orbit, I get between a 4 and 5 m/s drift when I'm at a kilometer or more in range.  All I need the autopilot to do is put the prograde vector on the target marker, just like docking, only at really high speeds.  Any thoughts on using a docking autopilot to not care about approach speed and just point me in the right direction?

[nascarlaser1]

if you want big explosions BDarmory might help, but I am not sure since I don't use it .

[NFunky]

Yeah, I actually originally started out by trying to use BDArmory's AAMs in vacuum, and learned my lesson.  So I started messing around with modular missiles, but since the guidance wasn't any better than the stock SAS at hitting the target, I got curios about other options.  The "Small High Explosive Warhead" works spectacularly by overheating the part(s) it collides with.  You can even manually detonate it with an action group, but as noted above in this thread, it does no damage unless it makes contact.

I've been having some good luck putting verner thrusters on the missiles instead of RCS.  They are way more powerful, and I can now just translate until prograde is on target instead of having to rotate the whole missile.

Somewhat off topic, but I really think there needs to be a way to "look" around with the navball.  It's fixed field ov view means that I can't see my prograde vector and target sometime when having to do violent maneuvers.  If I could look the navball on target, I could maneuver more precisely.