Hi, I am working on an optimization problem, which need to be solved periodically. I am presently solving it using analytical methods. It limits my problem definition. But if I use numerical or heuristic methods it will consume time. So whether anybody have any idea of solving an optimization problem recursively.
I want to optimise an objective function for a control problem at very short sampling instant. The optimisation problem will get updated at every instant with the present measurements from the process. The computational demand makes it difficult to get the solution with in this small duration. So if I could use the previous solution with the present measurement to find the optimisation solution, I can save computational effort and time.
Hey Lalu, Can you elaborate what do you mean by recursively here?? As far as I know heuristics methods such as GA, PSO, DE etc. are recursive in nature. All updates their solutions recursively. Yes, you are right heuristics methods take time. But you can set iteration limit as required or error limit.
If you want a fast implementation recursive is not the best way to go. If you could elaborate better we could help you more. Most heuristics are fast and can be implemented iteratively. But if you want it to be really fast it should probably be problem-specific.
I want to optimise an objective function for a control problem at very short sampling instant. The optimisation problem will get updated at every instant with the present measurements from the process. The computational demand makes it difficult to get the solution with in this small duration. So if I could use the previous solution with the present measurement to find the optimisation solution, I can save computational effort and time.
Perhaps a randomized approach may help?
http://en.m.wikipedia.org/wiki/Las_Vegas_algorithm
- Badges
- Science topic