Abstract: Air-assisted centralized metering can be operated at high speed for rice, wheat, and rapeseed. However, the seed stirring device is prone to damage the seeds, leading to the low stability of discharge quantity. In this study, a spiral-inclined flexible seed-stirring device was designed for high-speed, air-assisted, and centralized metering. The influencing factors on the seed damage were determined using Hertz contact theory, including the material properties of the seed stirring rod, the top structure, and the stirring speed. A comparison was made on the variation in the top structure of the seed stirring rod and the contact force of the population over time using EDEM simulation. A systematic investigation was also implemented to clarify the impact of the seed stirring rod on the seed supply. The test results show that the tangential and normal forces of the arc-shaped seed stirring rod in contact with the population were smaller than the trapezoidal seed stirring rod. There was better stability of seed supply when the inclination angle of the seed stirring rod was 45°. A single-factor test was conducted on a bench to determine the optimal range of seed stirring speed ratio. The results showed that the rice and wheat seeds performed better within the range of 1-2 seed stirring speed ratios, while the rapeseed seeds were in the range of 0.5-1.5 seed stirring speed ratios. A three-factor, three-level, and quadratic rotation orthogonal experiment was carried out to optimize the influencing factors. A regression model was then established for the damage rate, seed supply rate, and stability coefficient of variation of rice, wheat, and rapeseed. There was an optimal matching relationship between the stirring speed and the operating speed within the range of 10-14 km/h. Field experiments were also conducted to verify the simulation. It was found that the stability coefficients of the variation in the total displacement of rice, wheat, and rapeseed were 1.92%, 1.27%, and 1.14%, respectively, under the optimal combination of parameters, when the operating speed was 12km/h. The coefficients of variation in the total number of seedlings within 1 m² were 15.47%, 12.98%, and 17.93%, respectively, which fully met the requirements of the standard for the high-speed seeding. The finding can provide a strong reference to achieve high-speed and low-loss seeding in the parameter selection of high-speed, air-assisted, and centralized metering devices for rice, wheat, and rapeseed.