In production environments involving human-robot collaboration, effective management of motion-induced fatigue is critical for optimizing productivity and safeguarding worker well-being. Human movements in collaborative industrial tasks are inherently complex, shaped by repetitive task execution, time constraints, and continuous physical interaction with robotic systems. These characteristics motivate the need for robust methods capable of predicting and mitigating human physical fatigue within dynamic and fast-paced collaborative workflows. Autonomous generation of postural data, independent of body instrumentation and physical data collection setups, offers a compelling alternative for addressing these challenges. This work adopts an integrated approach that combines physics-based simulation with visual pose-based motion analysis to evaluate fatigue in upper-limbs movements during collaborative tasks. CAD-based human and robot models are employed to generate representative motion primitives, onto which fatigue coefficients and temporal decay models are applied to emulate progressive degradation in upper-limb motion performance. The resulting fatigue-modulated motion patterns are encoded using posture-based representation and dynamic movement primitives. The fatigue-induced motion patterns are deployed within a ROS2_RViz simulation environment, where they serve as inputs for robot tracking and adaptive task pacing. By capturing deviations in motion kinematics and timing across repeated task cycles, the proposed framework enables real-time fatigue estimation and supports adaptive adjustment in human-robot task coordination.