The parallel structure has advantages over the chain structure in that it has a higher rigidity, a greater power-to-weight ratio, but it also has a smaller workspace and less flexibility than the chain robot. Due to the small working space and limited dexterity, when designing robots in parallel, criteria such as firmness, minimum velocity transmission coefficient, and dexterity must be maximized to compensate for the above limitations. In many published methods of parallel robot optimization design, the Atlas method shows many advantages. But it has limited due to the interfering with the subjective judgment of the designer. To overcome this, the author proposes a new design method to improve the atlas method without implementing graphics as the original theory but instead using the spread-spectrum technique proposed by the author. The results of the paper show that it can be applied to parallel robots that have up to 4 design parameters and are highly general.