煤矿掘进机精准定位是智能掘进的基础，但井下低光照、高粉尘等恶劣作业环境导致单一定位方法精度低、稳定性差。为提高掘进机在恶劣环境中的定位精度，提出了一种基于误差状态卡尔曼滤波（ESKF）的激光雷达与惯导融合的掘进机定位方法。首先，以悬挂在巷道顶部的球靶中心为巷道坐标系原点，设计基于密度的噪声鲁棒空间聚类（DBSCAN）算法和基于形状特征的球靶点云提取算法，解决传统依靠反射强度区分球靶的方法在粉尘堆积时易失效的问题，结合坐标变换方法构建雷达位置测量系统以获得融合定位基准。其次，利用惯导积分得到掘进机的位置和姿态信息。然后，基于一阶高斯马尔可夫过程进行误差状态建模，采用误差状态卡尔曼滤波算法融合雷达和惯导的输出，得到掘进机在巷道中的融合定位结果，并将融合定位结果反馈给惯导，以校正其累计误差，从而获得精准的定位结果。定位试验结果表明：在掘进机静止状态下，不同位置和姿态角下雷达定位系统的位置误差小于10 cm，惯导定位系统的位置误差小于70 cm；在掘进机运动状态下，融合系统的位置误差为5.8 cm，相比雷达系统的位置误差降低了12.1%。基于激光雷达与惯导融合的掘进机定位方法可以在复杂掘进工况中满足煤矿掘进机自动截割时的定位需求。

Accurate positioning of roadheaders in coal mines is fundamental to intelligent tunneling. However, harsh working conditions, such as low illumination and high dust levels in underground mines, often degrade the accuracy and stability of single-source positioning methods. To improve the positioning accuracy of the roadheaders in these harsh conditions, a new positioning method based on the fusion of LiDAR and inertial navigation using error state kalman filter (ESKF) was developed. First, the center of the spherical target suspended from the tunnel roof was defined as the origin of the tunnel coordinate system. A density-based spatial clustering of applications with noise (DBSCAN) and a shape-feature-based spherical target point cloud extraction algorithm were designed to address the problem that conventional methods relying on reflection intensity for distinguishing spherical targets fail in environments with dust accumulation. The coordinate transformation method is then used to build a radar position measurement system to obtain a reference for the fusion positioning. Next, position and attitude information of the roadheader were obtained through inertial navigation integration. Subsequently, an error-state model was formulated based on a first-order Gaussian-Markov process, and the ESKF algorithm was applied to fuse the outputs of LiDAR and the inertial navigation, providing the fusion positioning results of the roadheader within the tunnel. The fusion positioning results were then fed back into the inertial navigation to correct accumulated errors, achieving precise positioning. Experimental results demonstrated that, under static conditions, the position error of the LiDAR-based positioning system remained below 10 cm across different positions and attitude angles, and the inertial navigation system exhibited a position error of less than 70 cm. In dynamic conditions, the fusion positioning system achieved a position error of 5.8 cm, reducing the LiDAR system's error by 12.1%. The proposed LiDAR and inertial navigation fusion-based roadheader positioning method meets the positioning requirements for automated cutting operations of roadheaders in complex tunneling conditions.