An initial command oil pressure of a lock-up control of coasting lock-up region is set by using a lock-up learning correction amount calculated by a learning control in a lockup control of the other of the plurality of lock-up regions such as slip lock-up region and complete lock-up region when one of the lock-up control of these other lock-up regions is performed before the lock-up control of the coasting lock-up region. Thus, a lock-up learning correction amount can be commonly used among three kinds of lock-up regions, and a lock-up as set is early realized in each lock-up region.