针对目前由于综采工作面工况恶劣，本安型巡检机器人存在通过性差、防护能力弱、感知技术手段有限，不能实现常态化运行等问题，提出隔爆巡检机器人方案，以应对综采工作面严苛工作环境。分析了轨道式巡检机器人在该环境下遇到的3种典型的工况：爬坡、错切、扭转。基于轨道式巡检机器人遇到的3种典型工况，设计了一种具有多自由度的隔爆巡检机器人，该巡检机器人除具有视频感知手段外，还增加了激光slam实时建图能力，能提供精确的三维感知能力，并且其感知能力不受光照条件的影响，能在无光条件下使用。针对爬坡工况，采用履带式行走机构；针对错切工况，设计了具有前后2组差速驱动的行走机构，其具有主动转向功能；针对扭转工况，在车身中部增加了纵向转轴。为了对其进行有效的控制，确保其在综采工作面平稳运行，充分考虑其具有约束的特点，分析得出其具有轨迹及底盘约束的运动学模型。同时介绍了该巡检机器人的控制方法，依据激光slam所建的实时三维激光点云地图，提取参考轨迹，根据参考轨迹实时计算各个履带主动轮的速度给定。基于运动学模型，提出一种模型预测控制算法，采用主从控制策略，模型预测控制只在前组驱动实现，后组驱动的控制量根据前组的速度给定及约束计算得出。该控制算法在计算复杂度及实时性方面做了平衡，控制算法基于ROS系统实现。最后，通过充分的井下工业性试验，验证了该控制方法能有效地解决爬坡、错切、扭转等综采工作面的苛刻工况，能在综采工作面执行常态化巡检任务。

Aiming at severe working conditions of fully mechanized mining face, poor passage, weak protective ability, and limited means of perception technology exists for intrinsic safety inspection robot and cann't be normally operated. The explosion proof inspection robot scheme is proposed for severe working conditions. And three characteristic working conditions: climbing the hill, shear, torsion are analysized for railed inspection robot of mechanized mining face. Based on three kinds of harsh working conditions, a kind of multiple degrees of freedom explosion-proof inspection robot is designed. Besides the video perception method, the lidar slam mapping ability is added, then, the precise 3D perception capability is provided. And the perception is not affected by lighting conditions, so it is can run in lightless conditions. For climbing the hill working condition, tracked walking mechanism is adopted. For shear working condition, a kind of walking mechanism which has front and rear differential drives are designed and can steer on its own. For torsion working condition, longitudinal rotated axis is added. In order to control this inspection robot effectively and ensure steady running in working face, the kinematics model is analysized and gotten aiming at railed mechanized mining face inspection robot, and is subjected to trajectory and chassis constraints. The control method of the inspection is introduced. The reference trajectory is extracted from the point cloud which is obtained from slam. The velocity of every tracked active wheel is calculated according to the reference trajectory. A kind of MPC control method is proposed based on the kinematics model, and the MPC is implemented only for front differential drive, the control setpoint of rear drive is calculated by the speed reference of front differential drive and the constraint between them. This method balances between the computational complexity and real time performance. And the algorithm is realized based on ROS. Lastly, through enough underground industrial tests, this method is effective for climbing the hill, shear, torsion working condition and inspection tasks can be executed by normalization.