Abstract: Polyphase sequences which have a low autocorrelation ensure an easily detectable peak in the output of a matched filter of a radar receiver. The low autocorrelation for polyphase codes is usually described in terms of the maximum magnitude of its sidelobes level. In this paper, an evolutionary algorithm combined with a local optimizer is used to search for polyphase codes with a small sidelobe level of an aperiodic autocorrelation function. The evolutionary algorithm is based on a floating-point representation and the Gaussian mutation is used to produce offspring for the next generation. The local optimizer is applied to find polyphase sequences which are good starting points for the evolutionary algorithm. This research demonstrates that optimization methods can effectively find polyphase codes with the low autocorrelation and seems to be very promising for future research in area of computer optimization for radar polyphase codes synthesis.