野外地面移动机器人实时环境建模与运动规划
文献类型：学位论文
| 作者 | 陈成
|
| 学位类别 | 博士 |
| 答辩日期 | 2015-05-26 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 中国科学院沈阳自动化研究所 |
| 导师 | 韩建达 |
| 关键词 | 野外地面移动机器人 实时环境建模 运动规划 局部轨迹规划 全局路径规划 |
| 其他题名 | Real-time Environment Modeling and Motion Planning for Outdoor Ground Mobile Robots |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 作为机器人领域中的一个重要分支，地面自主移动机器人在近几年来越来越多地应用于工业、家庭等典型的室内结构化环境。随着人工智能、物联网和大数据等学科的不断发展，地面自主移动机器人的应用领域也进一步扩展至以外星球、战场、极地等为代表的野外非结构化环境。在移动机器人作业环境不断扩展的同时，人们对其自主行为能力的要求也越来越高。作为地面移动机器人自主行为的基本要素，环境建模和运动规划也在近年来吸引了越来越多的注意。本文主要结合野外环境下地面移动机器人自主导航需求开展如下研究：对于野外环境，由于受量测数据误差的影响，现有的环境建模方法难以给出有效的环境描述方法以及可靠的障碍物检测算法。为了较为精确地构建机器人周边环境模型，本文提出了相对概率栅格高程图及其更新算法。进一步地，本文基于该环境模型提出了对应的障碍物检测算法。对于移动机器人运动规划，最基本的一个问题是如何生成一条连接局部初始状态及目标状态的可行轨迹，即局部轨迹规划。为了保证规划的轨迹是完全可行的，本文基于4阶贝塞尔曲线提出了一种新的可行轨迹规划方法。该方法将轨迹规划问题分解为轨形规划及速度规划两个子问题。轨形规划主要研究如何生成一条满足机器人初始及目标状态约束、运动学约束、转向连续有界约束的轨线。速度规划则主要研究如何在满足速度连续、加速度连续有界且不会发生侧向滑动的速度以跟踪执行该轨迹。对于移动机器人运动规划，另一个问题是如何在已知地图的情况下生成一条连接全局初始状态及目标状态的可行路径，即全局路径规划。为了生成运动学可行的全局路径，本文基于轨迹规划结果提出了偏差航向角轨迹最短的最小运动基元生成方法，并将生成的最小运动基元应用于构建启发式查找表及搜索算法以规划出运动学可行的全局路径。本文进一步地对以上3个方面研究结果进行离线仿真并将其应用于野外环境下的无人驾驶汽车及极地机器人，仿真及实验结果都验证了本文提出算法的有效性和实用性。 |
| 索取号 | TP242/C44/2015 |
| 英文摘要 | As one of the important branches of robots, ground mobile robots have been extensively used in indoor structured environments such as industry and families. Along with the development of artificial intelligence, Internet of Things and big data, ground mobile robots have been further deployed to work at outdoor unstructured environments such as alien planets, battle fields and polar regions, etc. Meanwhile, higher requirements have also been imposed on the autonomy of mobile robots. As two basic elements of the autonomy of mobile robots, environment modelling and motion planning have drawn more and more attentions in recent years. With the hope to improve the autonomy of outdoor ground mobile robots, this dissertation conducts researches in following areas:Affected by the uncertainty of measurement data collected by outdoor mobile robots, existing environment modelling algorithms could not effectively depict the surrounding environment and reliably detect hazards. To accurately construct models of the surrounding environments, a relative probabilistic environment model and corresponding update policy are proposed. To reliably detect hazards, another obstacle detection algorithm based on the model is also presented.For mobile robot motion planning, one basic problem is how to generate a feasible trajectory that connects the local start and target states, i.e. trajectory generation. To generate a feasible trajectory for the mobile robot, a trajectory generation algorithm based on quartic B$\acute{\text{e}}$zier curve is proposed. The algorithm decomposed the problem into generating the shape of the trajectory and generating velocity to track the trajectory. Trajectory shape generation mainly focused on generating a trajectory that satisfies the initial and target states constraints, kinematic constraints, continuous and bounded curvature constraints. Trajectory velocity generation mainly focused on generating a continuous velocity, continuous and bounded velocity and acceleration that could follow the trajectory while avoiding side-slip.For mobile robot motion planing, another problem is how to generate a path that connects the global initial and target states, i.e. global path planning. To plan a kinematic feasible path, based on the proposed trajectory generation algorithm, the dissertation has proposed an algorithm based on shortest trajectory of different heading angles to generate minimal motion primitives. The motion primitives are further applied to generate the heuristic lookup table and global path planning algorithm.The dissertation has further tested the proposed algorithms in simulation and experiments on an outdoor autonomous vehicle and polar rovers. The results have verified the effectiveness and practicability of the proposed algorithms. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 119页 |
| 源URL | [http://ir.sia.ac.cn/handle/173321/16809]  |
| 专题 | 沈阳自动化研究所_机器人学研究室 |
| 推荐引用方式 GB/T 7714 | 陈成. 野外地面移动机器人实时环境建模与运动规划[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2015. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。