Abstract: Aiming at the scheduling problem of roller grinding production line with multiple workpieces, multiple processes, multiple grinding machines and multiple cranes, an improved genetic algorithm is proposed to realize efficient and intelligent scheduling under the constraints of “workpiece-equipment-crane” space-time coupling. The scheduling model of the roller grinding production line is established, and the genetic algorithm is improved from four aspects of coding, selection, crossover, and mutation, so as to generate feasible solutions and improve the ability to seek optimization and jump out of the local optimum.Crane decoding rules and series operators based on action decomposition are further designed to complete intelligent scheduling. Taking the maximum completion time as the performance index and comparing with other algorithms, the results show that the proposed intelligent scheduling method is obviously better.