MechJeb2

MechJeb2

Mods
4M Downloads

Improve Principia node execution (fix multiple nodes bug)

DanielFabian opened this issue · 1 comments

commented

When using Principia with multiple nodes in the flight plan, sometimes, Mechjeb can screw up the burn.

The Principia timing on the active engine can occasionally be off and so MJ needs to kill rot in case the node was too short. It does that correctly if the node just disappears, i.e. in the case of the last node in the flight plan.

However, in the case of further nodes, principia doesn't delete the node but replaces it with the next one. Then Mechjeb gets confused, cuts the engine, re-aligns itself with the next node and makes a 2nd burn for the rest of the original Δv. This both underburns and wastes another ignition.

The root cause, I believe is this code snippet:

MechJeb2/MechJeb2/MechJebModuleNodeExecutor.cs

Lines 162 to 174 in 11dee5d

ManeuverNode node = hasNodes ? vessel.patchedConicSolver.maneuverNodes[0] : null;
if (hasNodes)
{
core.attitude.attitudeTo(Vector3d.forward,AttitudeReference.MANEUVER_NODE_COT,this);
}
else
{
if (!core.attitude.attitudeKILLROT)
{
core.attitude.attitudeTo(UnityEngine.Quaternion.LookRotation(vessel.GetTransform().up,-vessel.GetTransform().forward),AttitudeReference.INERTIAL,this);
core.attitude.attitudeKILLROT = true;
}
}

Specifically, line 162 should have another guard, according to: mockingbirdnest/Principia#2324 (comment)

This isn't a problem whatsoever on kOS. You need to lock steering to the moral equivalent of,

function node_steering {
  if not HASNODE
    return SHIP:FACING:FOREVECTOR.
  return NEXTNODE:DELTAV:DIRECTION
}

Once you start the burn,

wait until not HASNODE or NEXTNODE:ETA > 1.

The second part of this condition corresponds to having more than one maneuver stored.

This makes the burn script significantly simpler than the equivalent not using Principia.

So, I believe, we shouldn't be assigning node to maneuverNodes[0] if the ETA is in the future.

commented

closed by #1796