在煤矿智能化综采工作面的研究与应用中，液压支架的位置偏移已经成为制约综采工作面高产、高效、安全生产的难题。因此，深入研究液压支架在底板平面内位置偏移的检测与纠偏问题，对提升综采工作面智能化水平具有非常重要的意义。 针对综采工作面液压支架的位置偏移问题，分析了液压支架发生位置偏移的机理，建立了单台液压支架的位置偏移模型，将综采工作面液压支架的位置偏移分为前后未对齐偏移和不垂直于刮板输送机偏移两种基本的偏移类型，对单台液压支架的位置偏移进行了定义，提出了液压支架群位置偏移的检测与纠偏方案。为液压支架位置偏移的检测和纠偏奠定了基础。 针对液压支架位置检测基准的问题，研究了基于惯性导航技术的采煤机姿态的液压支架检测基准测定方法；提出了一种确定性误差的补偿方法，并对惯性器件的确定性误差进行了补偿；建立了基于刮板输送机轨道约束的采煤机姿态角变化模型，并提出了基于模拟退火粒子群最小二乘支持向量机的姿态角变化模型参数估计方法。为液压支架位置检测提供了精确的检测基准。 针对综采工作面“三机”运动坐标系oxy平面内液压支架的位置检测问题，提出了基于多源信息融合的液压支架立柱中心位置检测方法；提出了载体坐标系内基于改进CURE聚类算法和偏最小二乘法的液压支架立柱的识别和定位方法；分析了采煤机姿态角、液压支架推杆推移量和综采工作面“三机”坐标系内液压支架位置的关系，建立了基于二阶粒子滤波方法的综采工作面“三机”运动坐标系内的液压支架位置检测模型。解决了综采工作面“三机”运动坐标系oxy平面内液压支架定位方法问题，保证了液压支架位置检测的准确性。 在单台液压支架位置偏移研究的基础上，分析了综采工作面“三机”运行的实际情况，提出了综采工作面液压支架群的位置偏移评价方法。将综采工作面液压支架的位置偏移归纳为6种情况，建立了液压支架群位置偏移的评价指标体系。对每一种偏移情况分别建立了交叉信息粒化TS模糊故障树理论的液压支架位置偏移评价模型。解决了综采工作面液压支架群位置偏移的评价问题，为液压支架位置偏移的纠偏提供了依据。 针对液压支架群位置偏移的纠偏问题，提出了基于MAXQ分层强化学习的液压支架群位置偏移的纠偏方法。分析了综采工作面生产中“三机”的多种割煤方式对综采工作面液压支架群位置偏移的影响，归纳了9种对液压支架群纠偏的“三机”基本动作，建立了基于强化学习纠偏方法中关于状态-动作对的奖励函数，提出了基于MAXQ分层强化学习的液压支架群位置偏移纠偏方法，有效解决了液压支架群纠偏问题。 最后，在本课题组研发的综采工作面虚拟现实智能控制实验平台上，对本文建立的模型和提出的方法进行了实验验证，同时通过对采集的综采工作面现场部分数据进行了分析，实验结果表明本文所建模型和提出方法的有效性。

With the research and application of intellectual fully mechanized mining face, the position displacement of hydraulic support has been an critical problem for high production, high efficiency and safety production. Recently, in the plane of coal seam floor, the problem of the position displacement of hydraulic support focus on before and after the alignment among the hydraulic support, but not considering the crowd of the hydraulic support and the position displacement of scraper conveyor. Therefore it is meaningful to study the detection and correction for the position displacement of hydraulic support in the plane of coal seam floor. In this paper, focusing on the position displacement of hydraulic support, the mechanism of action has been analyzed. The model of single hydraulic support has been built. While the style of the position displacement of hydraulic support has been divided into two parts based on before and after the alignment and out of the vertical of scraper conveyor. The position displacement of single hydraulic support has been defined, and the methods of the detection and correction have been proposed for the position displacement of hydraulic support crowd. The work is done in this paper, which could be foundation for the detection and correction of the position displacement of hydraulic support. In this paper, focusing on the detection benchmark of the position displacement of hydraulic support, the measurement method of detection benchmark has been studied. The compensation method of determinate error has been proposed, and could be used on the inertial components. The adjustable model of miner’s attitude angle based on the constraint of scraper conveyor’s orbit has been built. The estimation method of the miner’s attitude angle has been proposed based on the simulated annealing algorithm(SA) and least square support vector machine(SVM), which could provide a precise inspection benchmark for the position displacement of hydraulic support. In this paper, the problem of the position detection of the hydraulic support is studied in the x-y plane of the "three machine" coordinate system at the fully mechanized coal face. The method for detecting the center position of the hydraulic support column is proposed based on multi-source information fusion. Proposing The method of identifying and locating the hydraulic support pillars is proposed based on the improved CURE clustering algorithm and least square method. The position detecting model of the hydraulic support in the “three machine” coordinate of fully mechanized coal face is established, which is based on the 2-order particle filter method. The problem of the positioning method of the hydraulic support could be solved in the oxy plane of the “three machine” coordinate in the fully mechanized coal face, which also could ensured the accuracy of the hydraulic support position detection. In this paper, the actual situation of the "three machine" motion of the fully mechanized mining face is analyzed, which is based on the study of the position deviation of single hydraulic support. The method for evaluating the position deviation of the hydraulic support group is proposed in the fully mechanized coal face. The position deviation of hydraulic support is divided into 6 classes in fully mechanized coal face. The evaluation index system is established for the position deviation of hydraulic support group. For each one of the classes, the deviation evaluation model of the hydraulic support group is established, which is based on TS fuzzy fault tree theory. The evaluation problem of the position deviation of the hydraulic support group is solved in the fully mechanized coal face, which could provide the required environmental information for the correction of hydraulic support deviation. In this paper, the problem of correcting the position deviation of the hydraulic support group is studied, while the method of correcting the position deviation of the hydraulic support group is proposed, which is based on MAXQ hierarchical reinforcement learning. With analyzing the influence of three kinds of coal cutting modes on the position deviation of hydraulic supports in fully mechanized coal face, the method for correcting the position deviation of hydraulic support group is proposed, which is based on MAXQ hierarchical reinforcement learning. All the methods are proposed in this paper, which could avoid the position deviation of hydraulic support group in the fully mechanized face, and ensure efficient production of the fully mechanized coal face continuously. Finally, based on the experimental platform of fully mechanized coal face "three machines", the models and methods are verified by a series of experiments. With the data analyzing, which is obtained from real fully mechanized working face, the models and methods is showed effectively.