在我国，煤炭资源是主要能源来源之一，但煤矿开采仍然是高危行业之一，无人化采矿旨在引入自动化和无人化设备，如自动化采矿设备、无人驾驶车辆和机器人等，来提高煤矿开采的安全性以及开采效率。运动规划是无人驾驶车辆和机器人实现自主导航和安全运作的核心技术之一，是实现机器人自主移动的有力保障，因此文本针对井下移动机器人的设计特点以及计算能力，围绕运动规划中的路径规划、轨迹优化、局部避障与路径跟踪三个关键问题开展相关研究，其主要内容如下：

针对煤矿井下非结构化、巷道狭长的地形环境以及传统A*算法探索时间慢，路径拐点过多，以及未考虑运动学，避障效果差的问题，提出了一个基于改进A*算法的前端路径搜索算法。通过限定A*算法搜索空间，优化A*算法的启发式函数及节点探索方向，引入跳点搜索方式，减少路径拐点以及探索时间。仿真结果表明相对于传统A*算法，改进之后的算法探索时间减少36.7%，探索点数减少33.6%，路径长度减少2.5%，弯折次数减少45.4%

针对井下障碍物众多，移动机器人后端轨迹优化计算量大，导致轨迹优化时间分配不合理，轨迹安全性较差，优化效率慢的问题，提出了一个基于STC（移动安全走廊）的轨迹优化算法，通过构建STC，优化碰撞检测范围，极大减少了计算压力。利用Bernstein多项式将轨迹曲线分解为多端连续的Bezier曲线，利用比例因子s对时间进行重分配，将轨迹的控制点以及高阶动力学约束在STC内部，保证轨迹优化的安全性。仿真结果表明，本文算法在目标成本上下降20.2%，轨迹长度上减少3.55%，运算时间上减少20%，移动速度上加快7.46%。

针对机器人感知信息有限以及先验地图信息出现变化状态下的动态轨迹规划问题，通过构建基于传感器信息的欧式体素距离（ESDF）地图来进行轨迹优化，通过改进Ego-planner算法，实现了移动机器人在线构建局部地图，躲避绕行障碍物，利用MPC算法实现轨迹跟踪，实验结果表明，移动机器人可有效躲避障物且轨迹跟踪误差满足实际需求，误差范围在±6cm以内。最后在履带式机器人上，验证本文运动规划算法，分别在室内简单环境、室内复杂环境以及模拟煤矿环境下等验证中本文算法的有效性。

In China, coal resources are one of the main sources of energy, but coal mining remains a high-risk industry. The goal of unmanned mining is to introduce automation and unmanned technologies, such as automated mining equipment, driverless vehicles, and robots, to enhance the safety and efficiency of coal mining. Motion planning is a core technology for driverless vehicles and robots to achieve autonomous navigation and safe operation, providing robust support for their autonomous movement. Therefore, this thesis focuses on the design characteristics and computational requirements of underground mobile robots, and conducts research on three key issues in motion planning: path planning, trajectory optimization, local obstacle avoidance, and path tracking. The main contents are as follows:

Addressing the unstructured and narrow underground terrain environment in coal mines, as well as the issues of slow exploration time, excessive path inflection points, lack of kinematic consideration, and poor obstacle avoidance in traditional A* algorithms, an improved A* algorithm is proposed for front-end path search. By limiting the search space of the A* algorithm, optimizing its heuristic function and node exploration direction, and introducing a jump point search method, the number of path inflection points and exploration time are reduced. Experimental results show that compared to the traditional A* algorithm, the improved algorithm reduces exploration time by 36.7%, exploration points by 33.6%, path length by 2.5%, and bending times by 45.4%.

Given the numerous obstacles underground and the large computational burden for trajectory optimization in the backend of mobile robots, leading to unreasonable time allocation for trajectory optimization, poor trajectory safety, and slow optimization efficiency, a trajectory optimization algorithm based on Safe Trajectory Corridor (STC) is proposed. By constructing the STC and optimizing the collision detection range, the computational burden is significantly reduced. The trajectory curve is decomposed into multi-segment continuous Bezier curves using Bernstein polynomials. Time reallocation is performed using the scaling factor s, with the control points and high-order dynamic constraints of the trajectory contained within the STC, ensuring the safety of trajectory optimization. Simulation results show that the proposed algorithm achieves a 20.2% reduction in target cost, a 3.55% reduction in trajectory length, a 20% reduction in computation time, and a 7.46% increase in movement speed.

To address the problem of dynamic trajectory planning under limited sensor information and changes in prior map information, a trajectory optimization approach is implemented using a Euclidean Signed Distance Field (ESDF) map constructed from sensor information. By improving the Ego-planner algorithm, the mobile robot can construct a local map online, avoid obstacles, and utilize the Model Predictive Control (MPC) algorithm for trajectory tracking. Experimental results demonstrate that the mobile robot can effectively avoid obstacles and the trajectory tracking error meets practical requirements, with an error range within ±6cm. Finally, the effectiveness of the motion planning module in practical scenarios is validated on mobile robots through experiments in indoor simple environments, indoor complex environments, and coal mine environments.