4 ms·

I wonder what happened there. Looks at 4 seconds the guidance system figured it needed to pull an ultra manoeuvre to slow the the horizontal speed(was drifting

by Rapzid 12y ago

I wonder what happened there. Looks at 4 seconds the guidance system figured it needed to pull an ultra manoeuvre to slow the the horizontal speed(was drifting left) but ended up with a solution putting the rocket upright again at nearly 0m. If it would have left some altitude for stabilizing I think they would have had it. Arm chair rocket science is fun.

TeMPOraL 12y ago
According to Elon's tweet that seems to have disappeared by now, they had static friction in bipropellant valves introduce a lag that the control system couldn't handle.
- Rapzid 12y ago
  It made a solid effort.
  - TeMPOraL 12y ago
    That it did, and it almost succeeded.
- shostack 12y ago
  Can you or someone else please explain what that means in slightly less technical terms without dumbing it down too much?
  - jffry 12y ago
    So presumably this is a valve controlling fuel flow to the engine. The valve is hooked up to some form of motor/actuator, and the rocket's control system needs to open/close the valve to control some aspect of the thrust. Since valves cannot open instantaneously, the control system has to know how long the valve takes to open/close, and start that process a little bit early so the valve is open when it needs to be open. If something goes wrong with the valve, it doesn't open as quickly, and so the computer model thinks it should be open and isn't, which could easily go off into an unstable state.
  - TeMPOraL 12y ago
    A control system is a feedback loop that connects measurements (orientation, speed, acceleration) and steering (grid fins orientation, engine power and gimbaling, RCS/other means of attitude control) in order to steer the entire system into a desired state (in this case, upright orientation and 0m/s velocity exactly at the point of touchdown). The control system measures the state of the rocket and directly adjusts the control inputs - i.e. it's not a series of if-else instructions, but more like "engine_gimbal_x = P * yaw + I * integral(yaw) + D * dyaw/dt"[0] - the inputs and outputs are wired together through a set of mathematical transformations. The "lag" is a delay between your steering signal and the state of the system changing in response. In this case, because of static friction, valves controlling fuel flow reacted with delay (I'm not sure if they're talking about Merlin's valves or attitude control engines though) - so the rocket did not react immediately after being told to do so. There's a whole math field around that topic, known as "control theory". There are ways to determine the limits of inputs and outputs that will allow your control system to keep everything in the desired state. In this case however, the input lag wasn't accounted for and it pushed the system out of the space of stable states, beyond the ranges where the control unit could keep up with the changes, so it ended up oversteering. Fortunately, in this case it's mostly a software problem, i.e. they could tweak the controller to deal with the lag. The best way to imagine it is to recall your first experiences with a bike or ice skates - trying to balance yourself on them is exactly the kind of thing the rocket was doing, probably with similar results. My control theory is a bit rusty nowadays, but I hope this clears up some things. [0] - if you ever hear about "PID controller", this is exactly where the name comes from - proportional, integral, derivative.
    - shostack 12y ago
      This was exactly what I was looking for. Thanks for taking the time to go into this level of detail. So looks like if they can solve for modeling this latency (likely a software problem), there's a decent shot they'll nail it the next time around?
      - TeMPOraL 12y ago
        If this turns out to be the only problem then yes, they have a decent chance of doing a proper landing the next time around.
- gridspy 12y ago
  Based on this quote "static friction in bipropellant valves introduce a lag that the control system couldn't handle" It sounds like the fuel valves which control the engine's throttle took longer to turn up or down the rocket than expected. The control system that is constantly processing rocket position and turning that into throttle outputs couldn't handle how long it took for decisions to turn into actions. That tends to cause wobbling in most controllers (overshoot, undershoot, etc).
InclinedPlane 12y ago
As mentioned, it's control system lag, which led to over-correction oscillations. This is very straightforward to deal with.
- tsotha 12y ago
  I find this a bit surprising given the amount of testing they did with F9R.

by Rapzid 12y ago

TeMPOraL 12y ago

According to Elon's tweet that seems to have disappeared by now, they had static friction in bipropellant valves introduce a lag that the control system couldn't handle.

Rapzid 12y ago
It made a solid effort.
- TeMPOraL 12y ago
  That it did, and it almost succeeded.
shostack 12y ago
Can you or someone else please explain what that means in slightly less technical terms without dumbing it down too much?
- jffry 12y ago
  So presumably this is a valve controlling fuel flow to the engine. The valve is hooked up to some form of motor/actuator, and the rocket's control system needs to open/close the valve to control some aspect of the thrust. Since valves cannot open instantaneously, the control system has to know how long the valve takes to open/close, and start that process a little bit early so the valve is open when it needs to be open. If something goes wrong with the valve, it doesn't open as quickly, and so the computer model thinks it should be open and isn't, which could easily go off into an unstable state.
- TeMPOraL 12y ago
  A control system is a feedback loop that connects measurements (orientation, speed, acceleration) and steering (grid fins orientation, engine power and gimbaling, RCS/other means of attitude control) in order to steer the entire system into a desired state (in this case, upright orientation and 0m/s velocity exactly at the point of touchdown). The control system measures the state of the rocket and directly adjusts the control inputs - i.e. it's not a series of if-else instructions, but more like "engine_gimbal_x = P * yaw + I * integral(yaw) + D * dyaw/dt"[0] - the inputs and outputs are wired together through a set of mathematical transformations. The "lag" is a delay between your steering signal and the state of the system changing in response. In this case, because of static friction, valves controlling fuel flow reacted with delay (I'm not sure if they're talking about Merlin's valves or attitude control engines though) - so the rocket did not react immediately after being told to do so. There's a whole math field around that topic, known as "control theory". There are ways to determine the limits of inputs and outputs that will allow your control system to keep everything in the desired state. In this case however, the input lag wasn't accounted for and it pushed the system out of the space of stable states, beyond the ranges where the control unit could keep up with the changes, so it ended up oversteering. Fortunately, in this case it's mostly a software problem, i.e. they could tweak the controller to deal with the lag. The best way to imagine it is to recall your first experiences with a bike or ice skates - trying to balance yourself on them is exactly the kind of thing the rocket was doing, probably with similar results. My control theory is a bit rusty nowadays, but I hope this clears up some things. [0] - if you ever hear about "PID controller", this is exactly where the name comes from - proportional, integral, derivative.
  - shostack 12y ago
    This was exactly what I was looking for. Thanks for taking the time to go into this level of detail. So looks like if they can solve for modeling this latency (likely a software problem), there's a decent shot they'll nail it the next time around?
    - TeMPOraL 12y ago
      If this turns out to be the only problem then yes, they have a decent chance of doing a proper landing the next time around.
gridspy 12y ago
Based on this quote "static friction in bipropellant valves introduce a lag that the control system couldn't handle" It sounds like the fuel valves which control the engine's throttle took longer to turn up or down the rocket than expected. The control system that is constantly processing rocket position and turning that into throttle outputs couldn't handle how long it took for decisions to turn into actions. That tends to cause wobbling in most controllers (overshoot, undershoot, etc).

InclinedPlane 12y ago

As mentioned, it's control system lag, which led to over-correction oscillations. This is very straightforward to deal with.

tsotha 12y ago
I find this a bit surprising given the amount of testing they did with F9R.