Aiming at the stability of the output speed and the poor adaptability of the transmission system during the operation of a cotton picker, a control strategy of hydro-mechanical continuously variable transmission (HMCVT) for cotton pickers based on gray prediction and fuzzy PID is proposed. Firstly, the hardware and software of the existing hydraulic mechanical coupling transmission test-bed of cotton pickers are designed, and the HMCVT human-computer interaction measurement and control system is built by using LABVIEW 2020 software. Then, combined with the transmission theory, the control strategy and gray prediction model are designed. Finally, the continuity test, transmission efficiency test, and adaptive control verification test are carried out. The results show that as the input speed increases, the peak time of the pump motor output speed is prolonged, while the overall speed regulation process is smoother, and the output speed process of the HMCVT system is continuous. As the displacement ratio of the variable pump increases, the transmission efficiency of the hydraulic system increases accordingly, but the highest efficiency is around 0.8. At a working speed of 10 km/h, the transmission efficiency of the HMCVT system of the cotton picker is more than 80%, and the high efficiency of the mechanical system in the whole system makes up for the low efficiency of the hydraulic system, and the transmission efficiency of the cotton picker is the highest at 15−25 km/h. The speed under the adaptive control strategy is better, with good robustness to sudden torque changes and speed fluctuations in the range of ±0.0125% under external load conditions. This study provides a reference for future adaptive control of transmission output speed for heavy-duty vehicles and construction machinery.