The nature of the objective function makes it very difficult to derive an analytic derivative. Also, the objective is a function of the current values of PTZ, and hence changes in appearance every time the value of the PTZ vector changes, and every time a target enters the area being surveyed. I'm not sure if that makes it any clearer, haha.
Anyways, I will try out the algorithms and provide feedback on my results. Thank you for the help. Akshay. On Mon, Jun 24, 2013 at 2:31 PM, Steven G. Johnson <[email protected]>wrote: > > On Jun 24, 2013, at 3:27 PM, Akshay Morye <[email protected]> wrote: > > More information is as follows: > > - No. of unknowns: 3. These are the values of the pan angle, the tile > angle and the focal length in meters to be optimized. > - Differentiable: Yes. > - Analytical Derivative: Not available. > - Local Optimum sufficient: Yes. > - Non Linear constraints: Yes. The number of constraints increases > with the number of targets to be imaged. That being said, I want to start > off with a very simple scenario where there's 1 target, and 1 camera. > > > > With such a small number of variables, derivative-free methods (e.g. > COBYLA, or AUGLAG with BOBYQA) should be workable, although you could also > try something like SLSQP with finite-difference derivative approximations. > (Not sure why you can't take derivatives analytically.) > > Steven > > _______________________________________________ > NLopt-discuss mailing list > [email protected] > http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss > > -- Akshay A. Morye, Ph.D Candidate (Electrical Engineering), Control and Robotics Laboratory, UC Riverside. Ph: +1 951 941 3266 http://www.ee.ucr.edu/~amorye/
_______________________________________________ NLopt-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
