盾构在隧道掘进中其姿态调整及推进的控制是顺利掘进的重要保证。但是盾构所处地质环境尤其是局部地质环境的不确定性,使得其姿态调整及推进策略的制定成为一项重要挑战。本文在姿态调整及推进策略制定中引入分层强化学习方法来对盾构姿态调整及推进策略进行学习,进而实现最优策略。学习算法将MAXQ分层学习算法及SVM算法结合,较好的实现了盾构姿态调整及推进策略的学习。 Shield tunneling in the attitude adjustment and propulsion control is an important guarantee for smooth excavation. However, the uncertainty of the geologic environment in which the shield is located, especially in the local geological environment, poses an important challenge in terms of attitude adjustment and promotion strategy. In this paper, stratified reinforcement learning method is introduced in attitude adjustment and propulsion strategy development to learn the attitude adjustment and propulsion strategy of shield, and then to realize the optimal strategy. The learning algorithm combines MAXQ hierarchical learning algorithm and SVM algorithm to better realize the shield posture adjustment and propulsion strategy learning.