In this paper, we develop a decision-making framework for real-time production control considering the condition variation of robotic arms. Specifically, the temperature dynamics of robotic arms under different operation conditions is analyzed to assess the robotic arm's health status. Statistical models based on the observation of real-time information is firstly built to characterize the relationship between the robot temperature and time, considering various operation modes (i.e., capacity, working mode, speed). Then a loading process using the robotic arm is investigated and a continuous space Markov decision model is formulated to minimize the total processing time for a limited batch of products with different types. Numerical studies suggest that the performance of the proposed method is significantly better than the benchmark plans. Such a study reflects the necessity of joint consideration on the health condition of production assets together with production control, to maintain high productivity and utilization of the assets in production systems.