Nondiffracting beams maintain their intensity profiles over a large propagation distance without substantial diffraction and exhibit unique propagation trajectories, leading to scientific impacts in various fields. However, the nonlocalized intensity distribution of non-diffracting beams is restrictive for many practical applications. Thus, strategies to optimize the beam profiles remain much in demand. In this report, we demonstrate an evolutionary algorithmic framework for optical beam engineering and optimization and experimentally validate it by realizing quasi-nondiffracting radially self-accelerating (or self-rotating) beams in a high-resolution imaging system. The work reports a tightly confined side-lobe-suppressed helicon-like beam that largely maintains its properties of radial self-acceleration and non-diffraction in the 3-D space. The optimization method represents a new methodological avenue that can be extended to a broad range of beam engineering problems.