Control system for automatic transmission

A control method for controlling an automatic transmission, the automatic transmission including a friction engagement element that is engaged or released upon shifting of the automatic transmission, a solenoid valve that produces a solenoid pressure by application of a solenoid current, an engagement pressure control valve that receives the solenoid pressure as an operating signal pressure and controls an engagement pressure to be supplied to the friction engagement element according to the solenoid pressure, and an electronic control unit for controlling the solenoid current which includes a learning correction amount setting and storing section for setting and storing a learning correction amount that is obtained when shifting of the automatic transmission is experienced and an engagement pressure command value learning correction section for correcting an engagement pressure command value that is calculated according to a shifting of the automatic transmission transition state by using the learning correction amount, wherein the learning correction amount is divided into an initial learning amount that has no dependence upon the engagement pressure command value and converges into a constant amount owing to individual variation, and a time-dependent deterioration amount that has dependence upon the engagement pressure command value and varies owing to posterior deterioration of a control mechanism, the control method comprising the steps of:

• setting and storing a learning region correction amount and an initial learning amount which are obtained by a learning control in a learning region that is a limited input torque region;
• determining a time-dependent deterioration correction amount according to the engagement pressure command value and a degree of progress of time-dependent deterioration that is indicated by a difference between the learning region correction amount and the initial learning amount;
• and
• calculating a final engagement pressure command value as the learning correction amount by adding the initial learning amount at the time of correction to the time-dependent deterioration correction amount.