For the problem of obstacle avoidance trajectory planning of a robot arm, a robot arm obstacle avoidance method based on a genetic algorithm is proposed. It is based on the two problems that the motion process can avoid obstacles and the motion process is more stable and efficient. First, the motion of each joint is planned as a sixth-degree polynomial, and the coefficients of the sixth-degree term are set as the pending parameters, and the motion of each joint is changed by changing the pending parameters. Then, the fitness function is then constructed by calculating the collision detection, angular velocity limit detection, acceleration limit detection, and the total trajectory length and rotation angle for each joint. Finally, the fitness function is optimised using a genetic algorithm to obtain smooth, continuous, and collision-free trajectories. Matlab simulation experiments show that this method can obtain the optimal or suboptimal trajectory without collision.