Abstract: Abstract: Adaption control strategy is a core technology as one part of control software for automatic transmission. The purpose of such strategy development is to compensate deviations due to engine torque signal difference, tolerance of transmission parts during manufacture, purchasing, storage and assembly processes, characteristics change with mileage accumulation, and to ensure acceptable shift quality and good drivability for the whole life of products. Many efforts have been made aimed to design self-learning strategies especially for dual-clutch automatic transmission and automated mechanical transmission by establishing vehicle launch strategy, feedback pressure control and solenoid duty cycle adjustment. While due to lack of pressure sensors in hydraulic automatic transmission, most of theories cannot be used directly. In overseas, main introductions on adaptive methods for upshift and downshift are published, but key technical information is strictly protected by corresponding patents and no detailed knowledge for instance control procedure is issued. According to different shift phases such as filling phase, torque phase and speeding phase, separated adaption methods like self-learning approach based on torque estimation or turbine acceleration signal are analyzed. But pressure sensor signal, longitudinal acceleration signal or high engine torque resolution are required for the application of those strategies, which seems not suitable for series transmission or low-level vehicles with domestic brands. To meet the requirements for TCU (transmission control unit) control software of vehicle production, adaption function for shifting clutches is researched, and corresponding software structure is introduced which includes shift type decision, shift quality monitoring, parameter adjustment and complicated method for adaption data update and storage. In addition, according to different profiles in torque phase and speeding phase for power on upshift and power on downshift, adaption strategies based on shifting time and turbine speed flare issue are proposed respectively. Then, according to shift sequence and shift control parameter of self-learning approach, complete software control procedures for parameter detection, processing, calculation and response are illustrated. It is very helpful for further research since parameter self-adjustment procedure turns to be clear with step by step introduction. A patent testing method is produced for recording shift issues, shift points, driving cycles and checking adaption results. The vehicle testing results showed that through adaption parameter adjustment, shifting time achieves to the target value, 700 ms, during power on upshift with only 2 repeated operations and for power on downshift, engine and turbine speed flare can be avoided gradually which makes shift quality better and more solid. The shift shock is reduced to 5 m/s3, which achieves to production vehicle shift quality level. The deviations due to the inconsistence of transmission from vehicle to vehicle and from engine to engine during manufacture and assembly process and characteristics deterioration can be compensated in very short time according to current research outcomes.