Design of dielectric lens antennas by multi-objective optimization

Lens antennas are a promising device for realizing the anti-collision radars needed in intelligent transport systems (ITS) and multibeam antennas for satellite communication. They have simple structure, high gain and no feed blockage. Especially if the operating frequencies are above the microwave band, the resulting apertures with several tens of wavelengths, become practical. In advance ITS application, the antenna is expected to resolve several objects in azimuth plane. In satellite application, the terrestrial station should catch multiple satellites at various directions. These applications demand multibeam antennas. Typical antennas with multibeam attributes include the dual-reflector bifocal antenna (Rao, 1974). Another proposal is the bifocal lens (Brown, 1956), its multibeam characteristics has been evaluated (Peebles, 1988). Multibeam antennas have to form high gain and low sidelobe radiation patterns at different directions. The bifocal lens is guaranteed to equalize the aperture phase distribution on the specified design directions in the scanning plane. However, it is not guaranteed on the transverse plane due to its astigmatism, and no previous report has adequately addressed the design issues, especially the resulting degradation of the radiation pattern. Another multibeam lens antenna is the Luneburg lens (Luneburg, 1964). Though it offers many focal points at arbitrary directions, it has manufacturing problems, tapered dielectric constant, and heavy weight. Our solution is to propose an effective method that optimizes multibeam lens antennas by gain, beamwidth, and sidelobe level. Antenna designs based on GA are very attractive and various methods have been proposed (Altshuler & Linden, 1997; Jones & Joines, 1997); we have already proposed a Yagi-Uda antenna design based on the pareto-GA (Kuwahara, 2005). This chapter presents the design of a multibeam lens antenna based on the pareto-GA. The coordinates of the lens shape and the feed position are given as variables and are associated with GA chromosomes. From the variables, the radiation patterns are calculated. The values of the objective functions are evaluated from the radiation pattern. The objective functions are given as the gain and the sidelobe level on the scanning plane and the transverse plane. To balance these objective functions, we adopt the pareto-GA (Fonseca & Fleming, 1993). In the pareto-GA, individuals are ranked through multiple objective functions and selection 18