Simplex guided extremum seeking control for real-time optimization

This paper presents a novel approach to improve the global performance of extremum seeking control to optimize static target functions. This algorithm adopts the traditional extremum seeking control to find local extrema, and a direct search algorithm to search for the global extremum. The two methods are employed in an iterative switching manner. Theoretical analysis shows that under the proposed method, the state of the system converges to a known neighborhood of the optimal value measured during the run-time of the algorithm. Simulations and experiments demonstrate that this algorithm outperforms extremum seeking control or direct search algorithms for target functions with multiple extrema.