【作者】 张波涛；

【导师】 刘士荣；

【作者基本信息】 华东理工大学 ， 控制理论与控制工程， 2012， 博士

【摘要】 移动机械臂是一种兼具可操作性和可移动性的机器人,近10余年来在工业、军事、外太空等领域均受到与日俱增的关注。运动规划是移动机械臂执行任务的重要前提条件之一,其具体内容由任务所决定,受任务所限制。本文就移动机械臂运动规划领域内的若干关键理论与应用进行分析与研究,包括：移动平台的点到点运动规划、搜索任务中的路径规划、机械臂逆运动学求解等。本文的研究工作可以总结为以下几个方面：1.体系结构是用于定义机器人系统功能模块间的关系,确定任务执行的模式,是理论与技术的承载框架。在研究了典型体系结构的特点之后,针对移动机械臂及其任务的特点,设计了一种用于移动机械臂的混合式体系结构。以此基础,对关键技术进一步研究提供指导,以便有的放矢,并使该研究具有更强的可拓展性。2.移动平台路径规划是移动机械臂运动规划的基础之一。针对人工势场法存在的问题及势场参数选择对路径规划的影响,提出了一种用量子遗传算法优化势场参数提高路径品质、降低陷入局部极小概率的方法。同时,采取一种用均匀分布的电荷来描述不规则障碍物的方法克服障碍物描述困难的问题。以此为基础,进而构建了一种基于栅格-几何分层地图的路径规划方法,该方法通过人工势场法和改进A*法互补的方式来防止人工势场陷入局部极小,同时降低A’算法的计算量。3.在移动机械臂的搜索任务中,复杂地图大多存在回路。经研究发现在回路中不同的搜索方向可能导致不同的期望时间,在以往研究中该问题一直没得到关注。通过对该问题的研究,提出了一种回路方向决策方法,并将该方法与一种启发式路径规划方法进行结合以降低其期望时间,然后将该方法用于概率地图中的目标搜索。理论分析、仿真与实验均表明该回路方向决策法能提高原启发式搜索方法的性能,减少搜索所需的期望时间。4.针对移动机械臂搜索任务中的序列规划问题,提出了一种问题转化框架(MTF),以此建立中国邮递员问题(CPP)与多目标搜索问题的联系。借助于该框架,特征地图可以转换为一种标准拓扑地图,进而可以采用适于CPP的搜索算法来处理移动机器人多目标搜索问题。基于MTF的算法是一种最优算法,且具有与启发式算法近似的多项式复杂度。理论分析与实验表明基于该框架的算法能得到比其他几种算法更短的路径。5.针对冗余机械臂的全局路径规划问题,提出了一种通用的装载平面冗余机械臂的移动平台的全局规划方案。首先构造了该类型移动机械臂的运动学模型,然后采用第3章所介绍的改进人工势场进行移动平台的路径规划。一种改进的粒子群算法(PSO)则用于求解逆运动学。将该路径规划方法用于一种六连杆移动机械臂模型,并用一种自行构建的仿真平台进行测试。测试表明该方法可用于移动平台的路径规划和平面冗余机械臂的逆运动学求解。6.提出了一种基于生物信息的免疫克隆算法(ICABBI),并将其用于复杂机械臂的逆运动学求解。ICABBI将环境信息、免疫细胞历史信息和遗传信息引入了克隆选择机制。通过该机制,ICABBI实现了人工免疫系统内部的全局信息交互。用一些经典全局优化问题对该算法测试,测试表明该算法有较好的普适性与高维处理能力。将其用于一种六自由度冗余机械臂的逆运动学,测试表明该算法适于处理这种复杂的非线性全局优化问题,有较高的求解精度。更多还原

【Abstract】 As a kind of robot with manipulability and mobility, mobile manipulator system has drawn tremendous attention in many domains in the past decade, such as industry, military and deep space. Motion planning, decided and restricted by the task, is an important precondition for the applications of the mobile manipulator. This thesis focuses on several problems in the motion planning of the mobile manipulator, such as point to point motion planning of mobile platform, path planning in the target search task and the inverse kinematics.Contributions of this thesis can be summarized as follows:1. Architecture is used to define the relationships among function modules, decide the execution model of tasks, and works as a frame for carrying certain theories and technologies of mobile manipulator. In this study, several classical architectures are reviewed, and then, a hybird architecture is designed according to the characteristics of task executed by the mobile manipulator. This hybird architecture is employed to provide guidance for further detailed studies.2. Path planning for mobile platform is the foundation of motion planning for mobile manipulator. Considering the problems of artificial potential field (APF) and the influence of its parameters, a new version of APF is presented to improve the quality of the generated path and reduce the probability of appearing local minimum. This improvement is realized by optimizing several parameters using quantum genetic algorithm. To solve the matching problem between APF functions and obstacles, a novel method is proposed for describing obstacles, in which it is assumed that point charges distribute uniformly along the boundaries of obstacles. According to the above theories, a hierarchical path planning strategy based on grid-geometric map is proposed. In this planning strategy, the local minimum problem is avoided and computational consumption of A* is decreased by the cooperation between APF and improved A*.3. In the target search task with a mobile manipulator, there exist loops in most complicated environments. We found that different directions in loops lead to different expected-time, and no studies so far have been made to solve this problem. Therefore, a direction choosing method is presented, and applied to a heuristic algorithm to reduce its expected-time for target search. This improved heuristic algorithm is employed to plan paths for target search in probabilistic environments. Simulation and experiments demonstrated that this approach can reduce the expected time for target search.4. Considering the sequence planning problem in target search task of mobile manipulator, a novel map transformation framework (MTF) is proposed. With MTF, a feature map or a topological map can be converted into a standard topological map on which many graph search algorithms suitable for Chinese Postman Problem (CPP) can be employed to carry out path planning for multiple target search. Similar to some heuristic algorithms, the MTF-based algorithm is globally optimal and has polynomial order time complexity. Theoretical analysis and experiments all indicate that the route generated by the MTF-based algorithm is better than those planned by several other target search algorithms.5. To deal with the global motion planning of redundant mobile manipulator, a general method is proposed for the mobile platform with a planar manipulator. In this method, the model of this kind of mobile manipulator is constructed, the approach discussed in chapter 3 is applied to plan path for mobile platform, and an improved particle swarm optimization (PSO) algorithm is used to solve the inverse kinematics of manipulator. This general method is applied to do motion planning for a robot with a six-link manipulator, and then tested with a simulation platform built by us. Simulation shows that the proposed method can be used to plan a path for the robot and calculate inverse kinematics for this manipulator.6. A novel immune clonal algorithm, called the immune clonal algorithm based on biological information (ICABBI), is proposed and used to solve the inverse kinematics of complex manipulators. In ICABBI, the computational implementation of the clonal selection principle takes into account environmental information, cell history information and hereditary characteristics. This mechanism realizes information communication in whole artificial immune systems. We tested ICABBI with several classical global optimization problems to analyze its universality. Then, the ICABBI is applied to solve the inverse kinematics of a 6-DOF manipulator. ICABBI has some features that are unique among biology-based methods. It can deal with the some complex nonlinear global optimization problems, such as the inverse kinematics of complex manipulators, and has satisfactory performance on some high-dimensional problems.更多还原