This study delves into the radiation pattern synthesis of reconfigurable intelligent surfaces (RIS) / reflection metasurfaces. Through superimposing multiple single-reflection profiles, which comprise the amplitude and/or phase settings of all constituent elements, a single incident wave can be effectively reflected in multiple asymmetric directions. However, some mismatch and interference between adjacent reflection beams may be caused by this superposition as well. Additionally, it is constrained by the inherent limitation that achieving linear and continuous amplitude adjustments and phase shifts in real-world designs is challenging. Consequently, the reconfigurable amplitude and phase must be approximated to discrete values, necessitating the arrangement of reflection profile before and after optimization based on integer. Therefore, in this paper, we adapt the traditional particle swarm optimization (PSO) algorithm to discretized integer-based PSO by proposing the concepts of 'discard rate' and 'knowledge.' With the enhancement of the integer-based programming, the multiple asymmetric reflection pattern can be synthesized with suppressed sidelobe levels within limited iterations and time cost.