With the evolution of large−scale customized production, logistics scheduling in Flexible Manufacturing Systems (FMS) has become a critical factor in improving factory efficiency. In highly automated workshops, heterogeneous material handling subsystems—such as Automated Guided Vehicles (AGVs), overhead conveyors, and ground conveyor systems—face challenges including high−concurrency tasks, dynamic resource conflicts, and complex physical constraints. To address these issues, this paper proposes a Hierarchical Local−search Mixed−Integer Programming approach (Hierarchical Local−MIP, H−LMIP). First, the complex heterogeneous shop−floor environment is abstracted into a directed graph topology with throughput characteristics, enabling an accurate representation of the physical and logical relationships among processing stations, idle resources, and transportation links. Based on this modeling framework, a hierarchical multi−objective optimization model is constructed, where the primary objective is to maximize the number of on−time completed tasks, while secondary objectives include minimizing activation cost, reducing transportation energy consumption (i.e., low−carbon objective), and balancing system load. To overcome the computational challenges caused by large−scale discrete decision spaces, the proposed H−LMIP incorporates a two−stage dynamic constraint−guided mechanism. By integrating weighted scalarization with local search, the method enables efficient exploration and refined trade−offs among hierarchical objectives within limited computational time. Numerical experiments demonstrate that the proposed H−LMIP achieves better overall optimization performance than Local−MIP, LNS, and the classical NSGA−II, providing effective decision support for low−carbon and efficient operations in smart factories.