Abstract: To achieve arc-second level positioning accuracy for the air-floating turntable, an improved particle swarm optimization algorithm and back propagati (BP) on neural network are employed to optimize the drive parameters based on 20 sets of empirical parameters and their positioning accuracy data. This method is based on particle swarm-optimized neural networks, which enhance particle position randomness through chaotic mapping and introduce a Lévy flight strategy to prevent local optima. A pneumatic rotary table was constructed for a comparative experiment on drive parameter optimization. After empirical tuning, the positioning accuracy of the rotary table was ±6.91″. With optimized tuning, the positioning accuracy improved to ±2.27″, a 67.15% enhancement. Moreover, the repeatability accuracy improved from±5.99″ with empirical tuning to ±2.00″ after optimization, marking a 66.61% improvement. These results demonstrate that the proposed parameter optimization method effectively enhances positioning accuracy.