Quadratic programming hanging

[FelipeC]

I'm having an issue to use the quadratic programming VI on Labview. After a long debug I found out that a sub VI called "qp_find initial feasible solution" is hanging randomically depending on the input parameters and always increasing the processing time. Some times takes some hours to stop its inner loop.

As a workaround I disabled this VI and placed a LP block to get a initial solution. It's not the best solution but now it's working like a charm. QP VI is taking 2 to 4 ms to get the solution instead of 20 ms to infinite. Actualy the LP is taking more time now then QP block. Now it's stable enough to run it on CompactRIO.

Is there another way to find a initial feaseble solution? A correction of this VI would help too.

PS.: my problem has 16 variables, 5 equality constaints and 32 parameter bounds and I'm using LV 2011 SP1.

[erick.yamamoto]

Hello FelipeC

So you can post this vi. Because this vi is excluded, and i don't have access.

Att.

[altenbach]

Can you attach a input dataset that demonstrates the problem?

(Just place an indicator in the cluster entering the subVI, run the VI to populate the indicator, abort the VI, then change the indicator to a diagram constant. Copy it to the diagram of a new VI and attach it here.)

[FelipeC]

I got one dataset showing the problem. I placed a "check time" VI inside the inner loop of "check feasible solution" VI to overcome the hang of the code. But it still cannot get the initial solution many times.

I'll try to get more datasets but this one shows the problem while only three variables are in use, a very simplified snapshot of the problem.

Thanks.

[Markanbl]

Hello, 

I have a similar problem. The optimization done in Matlab is working properly but when translated to Labview either the quadprog doesn't find the optimum solution or this subVI hangs indefinetely. The inputs at the first iteration are exactly the same so the optimization should work in the same way. Does anyone know what is the issue with this VI?

Thank you!