Ships occasionally won't align to maneuver nodes

Question

Ships occasionally won't align to maneuver nodes

SpareBuffalo opened this issue 7 years ago · 9 comments

SpareBuffalo commented 7 years ago

Very occasionally a ship won't align either via MJ or via the stock autopilot. The angle in both cases I have experienced has been off by only a couple of degrees. In MJ the error is greater than the stock, I have never experienced an error of more than ten degrees. This results in rotation about an odd axis, and MJ won't auto-warp or execute the node. It is verified fixed by returning to main menu.

The photo shows the alignment error. The craft was rotating around the maneuver node axis without pointing the nose straight into the maneuver node.

brianmcnett · Answer 1 · 2017-12-01T12:45:39.000Z

I appear to experience this as well, also with the result that auto-warp does not engage when this occurs. I've found that manually rotating the craft to 0° roll will momentarily restore alignment usually enough to engage auto-warp. This remains true for build #762 although the error appears smaller than shown. I haven't experienced the node-execution failure. I'm attempting to reproduce for the purposes of collecting logs.

…

On Thu, Nov 30, 2017 at 10:44 PM, SpareBuffalo ***@***.***> wrote: Very occasionally a ship won't align either via MJ or via the stock autopilot. The angle in both cases I have experienced has been off by only a couple of degrees. In MJ the error is greater than the stock, I have never experienced an error of more than ten degrees. This results in rotation about an odd axis, and MJ won't auto-warp or execute the node. It is verified fixed by returning to main menu. The photo shows the alignment error. The craft was rotating around the maneuver node axis without pointing the nose straight into the maneuver node. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#959>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGLKpo4V_8VCQRAU8KMBRHrDlBak1KU1ks5s76A-gaJpZM4Qx7w2> .

lamont-granquist · Answer 2 · 2017-12-05T06:58:52.000Z

In the Attitude Adjustment controller try using the "HybridController" and in general it seems the PID needs more Integral and less deadband (i think).

SpareBuffalo · Answer 3 · 2017-12-14T01:09:31.000Z

Okay, well I haven't had it happen since I posted this. Couldn't find any setting about HybridController, but I changed the other two settings. Pardon my ignorance, but I'm not at all sure what they changed lol. Don't even know what Deadband (have an idea) or Integral are (math 🙄).

brianmcnett · Answer 4 · 2017-12-14T03:08:05.000Z

I find that switching to HybridController frequently does nothing at all (I've had this work all of one time and could not recreate the conditions). I tried reducing the deadband, but it is already down in the thousandths. I will admit to being fuzzy about the workings of PID controllers generally. but presuming the P, I, and D are represented by Kp, Ki, and Kd respectively... Ki is the integral. What are good limits for that value?

…

On Wed, Dec 13, 2017 at 5:09 PM, SpareBuffalo ***@***.***> wrote: Okay, well I haven't had it happen since I posted this. Couldn't find any setting about HybridController, but I changed the other two settings. Pardon my ignorance, but I'm not at all sure what they changed lol. Don't even know what Deadband (have an idea) or Integral are (math 🙄). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#959 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGLKpg8e9sZkweb3DzzfmRGuI27IdWoCks5tAHVNgaJpZM4Qx7w2> .

lamont-granquist · Answer 5 · 2017-12-14T05:42:24.000Z

Yeah, no tweaking is working for me either, but I haven't gotten the time to Debug.Log what is actually going on

lamont-granquist · Answer 6 · 2017-12-14T05:43:28.000Z

manually increasing warp with '.' should work BTW, its just annoying.

brianmcnett · Answer 7 · 2017-12-14T22:24:20.000Z

Manually toggling warp on, and immediately back off again does work. Autowarp will then engage every time. Annoying is right.

…

On Wed, Dec 13, 2017 at 9:43 PM, Lamont Granquist ***@***.***> wrote: manually increasing warp with '.' should work BTW, its just annoying. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#959 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGLKpgrxmqqicGZ1b-hudS-IydE28S2Mks5tALWBgaJpZM4Qx7w2> .

lamont-granquist · Answer 8 · 2017-12-19T02:11:52.000Z

                if ((core.attitude.attitudeAngleFromTarget() < 1 && core.vessel.angularMomentum.magnitude < 0.05) || (core.attitude.attitudeAngleFromTarget() < 10 && !MuUtils.PhysicsRunning()))

its almost certainly that line in the NodeExecutor. it should probably use angular velocity instead of momentum. dunno if you can find that off of Vessel or off of MJ's vesselState already, or if you need to just divide by mass there, and offhand IDK what a good value would be.

lamont-granquist · Answer 9 · 2018-12-24T19:25:16.000Z

fixed this to use angularVelocity, using the HybridController and tweaking MaxStoppingTime to 1.0 or so is recommended.

Share to