Summary
International Symposium on Antennas and Propagation
2008
Session Number:1IS01A
Session:
Number:1IS01A-1
Hamilton Algorithm and Overview of Novel Optimization Techniques
Jun Cheng, Takashi Ohira,
pp.-
Publication Date:2008/10/27
Online ISSN:2188-5079
Summary:
A single-port compact array antenna, i.e., electronically steerable parasitic array radiator (Espar) antenna [1], have shown the potential for application to wireless communications systems, and especially to mobile terminals. The (M+1)-element Espar antenna has only an active radiator connected to the receiver. The remaining M elements are parasitic. The antenna pattern is formed according to the values of the loaded reactance on these parasitic radiators. Because of the configuration of the Espar antenna, we face the following three difficulties [1][2] in the development of optimum algorithms: a) The signals on all elements cannot be observed. Only the single-port output can be observed. b) The RF currents on the elements are not independent but mutually coupled with each other. c) The single-port output is a highly nonlinear function of the variable reactances that includes the admittance matrix inverse. In addition, unlike digital beamforming antennas, conventional criteria such as MMSE (Minimum Mean Square Error) are useless for the optimization of the Espar antenna, since the amplitude of the antenna output is difficult to be adjusted [2]. In this paper, we give an overview of criteria and optimization algorithms for beamforming and design of the Espar antenna. The criteria are maximum power, maximum cross-correlation coefficient and maximum m-th order moment [1][2][3]. We describe the optimization algorithms including random search algorithm [4], gradient-based algorithm [2][5], and Hamilton algorithm [6][7][8]. For the optimization of the Espar antenna, the gradient-based algorithm converge fast but sometimes unwillingly fall into a local minimum depending upon the initial value for one of their parameters. On the other hand, the random search algorithm tolerate local-minimum problems but rather slow to reach the final goal. Hamiltonian algorithm intends to meet the two conflicting requirements, i.e., to be deterministic and to be free from local problems. The algorithm is especially expected to work effectively in case that the number of parameters are very large.