In multiproduct mixed flexible manufacturing cell, the number of automated guided vehicles (AGVs) and their routings will directly affect the cell’s efficiency. Considering the constraints such as processing time, batch requirement and facility location, a mathematical model was built aimed at minimizing the handling time of AGVs. To cooptimize the number of AGVs and their schedulings under different distributions of manufacturing tasks, an improved Memetic algorithm was proposed with the adjustment of coding and search mechanism, which effectively avoids the problem that conventional crossovers easily result in illegal solutions and the optimized AGV routing is also obtained. At last, a case study is illustrated to prove the validity of the proposed method.