The rapid advancement of Industry 4.0 has transformed manufacturing, giving rise to Flexible Smart Manufacturing Systems (FSMS) capable of adapting to fluctuating market demands and operational uncertainties—essential for achieving mass customization. However, conventional approaches that separate long-term planning from real-time scheduling struggle to meet the demands of modern manufacturing environments, particularly in adapting to frequent demand fluctuations, managing system complexity, and ensuring rapid responsiveness. To address this challenge, this paper presents a demand-driven hierarchical framework that integrates planning and scheduling for flexible smart manufacturing, enabled by a mobile, multi-skilled, robot-operated system. First, a novel system identification model is developed using behavioral cloning to extract essential system properties that inform decision-making. Next, a coupled dual-loop control structure is introduced: an outer planner loop optimizes robot configurations based on demand forecasts, while an inner scheduler loop dynamically adjusts robot assignments in response to unexpected disruptions. The control strategy leverages the System Property-Infused Multi-Agent Deep Deterministic Policy Gradient (P-MADDPG) algorithm, which integrates dynamic system properties to improve adaptability and decision-making in complex environments. Extensive experiments are carried out to demonstrate the framework’s effectiveness in adapting to frequently shifting demands, minimizing resource waste, and achieving superior performance with higher throughput compared to existing approaches, thereby providing a robust solution for integrated planning and scheduling in personalized production.