Collaborative Production Planning with Unknown Parameters using Model Predictive Control and Machine Learning

This paper aims at addressing the class of collaborative production planning problems, in which the regulator does not know certain private parameters of the participators. The accuracy of decision parameters is of key importance in production planning, which makes the manufacturing decisions biased or even inaccurate. To deal with this issue, this paper employs the technique of machine learning into the model predictive control (MPC) framework to propose a comprehensive algorithm. It applies machine learning to solve a regression problem for estimating the unknown parameter values, and this procedure is based on the historical data obtained during the participators' individual decision making process. With the help of an accurate estimate, the regulator can use MPC to make the optimal decisions to maximize its overall net profit. The proposed algorithm is further checked by a simulation case study to validate its estimation accuracy and decision effectiveness. Comparisons are also made with individual and pure MPC decisions to confirm its advantages for profit increasing.