The invention relates to a quadruped robot movement track control method based on a PSO-PD (Particle Swarm Optimization-PD) neural network. The quadruped robot movement track control method comprises the following steps: (1) calculating a target falling point of a robot on a movement track; (2) inputting the target falling point and a feedback body gravity center position into an input layer of the PSO-PD neural network, performing linear conversion, inputting into a first hidden layer, performing ratio calculation and differential operation, performing linear conversion, inputting into a second hidden layer, and acquiring instruction displacement in an x direction and a y direction, inputting into a third hidden layer, calculating an instruction displacement distance, instructing a body direction, inputting into a fifth layer, performing foot tract control and steering control on a quadruped robot, inputting into a sixth layer, adjusting the posture of the quadruped robot in case of disturbance, inputting into a seventh layer, calculating the body direction and the body gravity center position of the quadruped robot, feeding back the body direction to the fifth layer, and feeding back the body gravity center position to the input layer. The quadruped robot movement track control method has relatively good non-linear decoupling control capability, is precise in control and good in stability, and has relatively good anti-interference properties.