Current selective compliance assembly robot arm （SCARA） robots cannot meet the requirements for high speed and high precision in electronic chip manufacturing and other industries due to its long joint driving chain. For this problem， a novel SCARA robot is proposed， which adopts a direct drive mode rather than the traditional servo motor add reducer driving mode. Its structure is very simple because the intermediate transmission links are removed. Meanwhile， an optimal design method for optimizing the robot structural parameters is proposed to solve the stability problem of the robot end caused by the large torque ripple of the direct-driven joint. After the maximum motion range of SCARA robot is determined， the velocity evaluation function and the stiffness evaluation function are constructed respectively， which are taken as the adaptive functions to optimize the arm length by a multi-objective genetic algorithm. The experimental comparison before and after the arm length optimization shows that， the joint speed is up to 3.9 m/s before optimization， which can be greatly improved to 4.6 m/s after optimization. It is found by simulation that the deformation is reduced after optimization with the same terminal load. The repeated positioning accuracy is ±0.010 5 mm before optimization， which is superior the traditional SCARA robot， and it is further improved to ±0.009 mm after optimization. Through simulation and experimental results， the genetic algorithm optimization can effectively solve the stability problem of the SCARA robot end， and can significantly improve the motion speed of the robot joint.