毕业生知识库

      移动机器人良好的环境感知和导航能力是自主执行任务的前提，是高质量任务执行的关键一环，具有重要的理论研究意义和广泛的应用前景。本文针对移动机器人环境感知与导航控制开展研究，论文的主要内容如下：

      首先，介绍了研究背景和研究意义，从移动机器人环境表示模型、基于点云的地形描述和分类、移动机器人导航控制三个方面进行了现状综述，并对论文内容和结构做了介绍。

      其次，给出了多线激光雷达的标定方法，根据环境建模的需求和点云数据的特点，建立了基于三维点云的多高度混合栅格地图用于环境的表征。进一步地，进行了地面分割和栅格聚类，得到了一系列点云簇，为后续处理提供了基础。

      第三，基于多高度混合栅格地图，提出了集成高度指数、崎岖度和斜坡角度三种特征的地形描述与分类方法。将主成分分析用于斜坡地形判定，并提出了基于RANSAC和最小二乘的斜坡角度估计算法，进行了实验验证。实验结果表明所提方法能够有效地提取障碍、坑洼和斜坡地形，并能够获得较为准确和鲁棒的斜坡角度估计结果。

      第四，针对四足机器人自身高频振动的运行特点，坑洼不平的路面环境和动态障碍的突然干扰等挑战，研发了基于多线激光雷达的四足机器人领航员检测和跟随方法。借助反射强度辅以形体特征对领航员进行实时检测，并采用改进的向量场直方图算法躲避跟随过程中随时可能出现的动态障碍。该方法在实际路面环境下进行了实验验证，效果良好。

      第五，利用Cartographer算法建立了全局混合栅格地图，在此基础上，提出了一种能够应对动态环境的基于Bezier曲线的实时路径规划方法。对动态障碍的位置进行实时观测，预测了它的运动轨迹。采用概率的形式表征动态障碍影响的区域，缓解了因动态障碍造成的全局地图拖影问题。进而结合环境和自身约束，以曲线长度和所通过动态栅格的概率之和为评价指标选择最优的Bezier曲线。实验验证了所提方法的有效性。

      第六，基于全局混合栅格地图，提出了基于A*算法和分段Bezier曲线的实时路径规划方法。所提方法以A*路径为指引，同时利用Bezier曲线对A*路径进行拟合。面向复杂环境，以拟合指数为指标，结合环境和自身约束，提出了分段Bezier曲线的优化生成方法。将各段最优Bezier曲线依次首尾相连即为规划的路径，方法的有效性通过实验进行了验证。

      最后，对本文工作进行了总结，并指出了需要进一步开展的研究工作。

  Environmental perception and navigation ability is the basis for mobile robots to execute tasks autonomously, and it is crucial to execute tasks effectively. It is significant in both research and applications. This thesis focuses on the environmental perception and navigation control for mobile robots. The contents are as follows:

  Firstly, the research background and its significance of this thesis is given. The research development of environment description model, terrain description and classification, and navigation control for mobile robots is reviewed. The contents and structure of this thesis are also introduced.

  Secondly, the calibration method for 3D LiDAR is given. The multi-level hybrid grid map based on 3D point cloud is constructed for the environment description, and the requirements of environment modelling and the characteristics of the point cloud are considered. The clusters of point cloud are generated by ground segmentation and grids clustering, which provide the basis for further processing.

  Thirdly, on the basis of the multi-level hybrid grid map, the terrain description and classification method integrated with height index, roughness, slope angle is proposed. The principal component analysis is used to determine the slope terrain, and a slope angle estimation method based on RANSAC and Least Squares is proposed. The experiment results show that the proposed method can extract the terrains of obstacles, potholes, and slope with accurate and robust slope angle estimation results.

  Fourthly, the navigator detection and following method for quadruped robots based on 3D LiDAR is developed to deal with the challenges of high frequency vibration of the robot, uneven ground, and sudden disturbance of dynamic obstacles. The reflection intensity and the body features are adopted to detect the navigator with the improved vector field histogram algorithm for the collision avoidance of the dynamic obstacles. This method is verified in actual ground environments with a good performance.

  Fifthly, the global hybrid grid map based on the Cartographer algorithm is constructed. The real-time path planning method based on Bezier curve is proposed, which is used to deal with the dynamic environments. The location of the dynamic obstacle is observed in real time, and its trajectory is predicted. The area affected by the dynamic obstacle is described probabilistically, which can alleviate the smear problem in the global map. The optimized Bezier curve of the method is selected considering the criterions of curve length, and the sum of the probability of all dynamic grids related to the curve. The method is verified by the experiments. 

  Sixthly, a real-time path planning method based on A* algorithm and Bezier curve is proposed with the global hybrid grid map. The result of A* algorithm is used to guide the direction of Bezier curve, while the Bezier curve is used to fit the A* path. The optimized curve generation method is proposed to meet the requirements of fitting index with the consideration of constraints from environment and the robot itself. The set of the optimized curve for each path segment is regarded as the result of path planning for the mobile robot. The effectiveness of the proposed method is verified through the experiments.

  Finally, the conclusions are given and future work is addressed.