Suitable Optimization Technique using real-time data

I am optimizing the performance of my controller to control a complex system using some parameter of the controller. I have about 10 optimization parameter to be used to optimize the performance. I am using non-liniear optimization tool to get the optimal controller parameter (10 in total as I said). So it is a big 10 dimentional space. I don't have much constraint except upper bound and lower bound of the optimization controller parameters. But the cost function (to indicate the controller performance) is very complex as combination of tracking error, oscillation, overshoot, settling time etc. So my optimization technique will send diffeetn controller parameter and get the real time data after one control cycle. Use those data to calculate cost function. Based on the cost function, it will try to optimiza the cost function and get the best controller parameter. I am just wondering why optimization technique can be best for me to converge faster with resonable accuracy? Thanks in advance.

Answer 1

[Catalytic]

Because what else would you use?

[mumin chy]

I am trying fmincon , lsqnonlin, Particle swarm optimization, 'simulannealbnd', 'patternsearch' generic algorithm.

Answer 2

[Matt J]

I am trying fmincon , lsqnonlin, Particle swarm optimization, 'simulannealbnd', 'patternsearch' generic algorithm.

The answer seems deceptively obvious. Use whichever of these shows the best performance.