Lei Liu, Yimeng Li, Haoran Li, and Dongmei Wang
Manufacturing system, autonomous control, reinforcement learning method based on the value function, reinforcement learning method based on policy gradient, deep reinforcement learning
In view of the existing problems of manufacturing system in traditional vocational education, such as insufficient data acquisition ability, low autonomous learning ability, and lack of intelligent decision accuracy, deep reinforcement learning is applied to manufacturing system control in vocational education through comparative analysis, so as to optimise and solve the existing problems. The present invention uses radio frequency identification technology (RFID) for pairing, and processes the data model obtained by combining. Based on the processed data and the needs and requirements of the manufacturing system, combined with comparative analysis, the paper compared the reinforcement learning method based on the value function (RLM-VF) and the reinforcement learning method based on policy gradient (RLM-PG). Through the convergence speed and stability, it is found that the performance of RLM-PG model is better. Finally, the performance of the constructed model is compared with that of the traditional manufacturing system in vocational education. The average scores of students with independent learning ability in RLM-PG model increased by 7.34% and 13.04%, respectively, compared with RLM- VF and traditional model. The conversion rate of RLM-VF and traditional method (TM) based autonomous learning models was 4.02% and 5.38% lower than that of RLM-PG based autonomous learning models, respectively. Manufacturing systems in vocational education based on deep reinforcement learning can better cope with complex problems and changing environments, and have adaptability and stronger autonomous control performance.
Important Links:
Go Back
