【作者】 杨帆；

【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， 2011， 博士

【摘要】 在过去20余年的发展过程中,多移动机器人系统依据其时间、空间的分布性与高效性展现了其无可匹敌的魅力,在工业、军事、国防、生活和深空等应用领域均受到了广泛关注。多移动机器人协作在诸如流水线生产、编队巡逻、仓储运输、交通规划与环境探索等应用领域的研究中均显示出了极大的优势。而在一些具体的应用任务,如大型物体的运输、目标的抓捕、多传感器协作地图探索等研究中,要求机器人组成特定的队形并在任务过程中保持队形不变或改变队形执行任务,以此引发了多机器人编队控制的研究热潮,并成为多移动机器人协作与协调的关键技术之一。本文以协作大型物体运输为应用背景,依据从底层运动控制、运动规划到高层任务规划的思路,对多移动机器人的编队行为控制到编队规划最终到编队任务分配做了相关研究。本文的研究工作可以总结为以下几个方面：1)作为机器人运动与编队控制的基础,首先设计了机器人基本运动行为和基于服务的运动行为结构。在基于行为的控制方法研究中,以往主要是针对具体任务或实验进行设计并且多为上层的描述性语言,缺少基本运动行为的设计和对行为的可重复利用的研究。本文提出以机器人与编队中虚拟目标机器人之间的位姿误差作为驱动输入,驱动轮速度作为输出,设计移动机器人的基本运动行为,运用具有可重用性的面向服务架构(service-oriented architecture, SOA),设计了基本运动行为的结构。基于服务的运动行为结构,只需根据编队任务进行适当修改或扩展、组合就能满足新的情况或任务。所设计的基本行为和基于服务的运动行为结构的可行性和可重用性,在仿真环境和物理实验中得到了验证,并在多机器人编队形成、编队运动与协作运输任务的研究中得到应用。2)在基本运动行为的基础上,研究基于Leader-Follower的编队形成与运动规划。本文提出了基于竞标机制的编队时间最优规划方法。该方法将多机器人编队问题转换成任务分配问题,即处于跟随位置的移动机器人对编队中的虚拟机器人位置进行投标,而处于领航者位置的机器人根据投标进行集中规划,将分布式投标方法与集中式优化方法相结合,在降低集中规划方法计算维数的同时,完成编队位置的指派任务。此外,利用竞标时产生的实时监督机制,实现在编队中机器人故障或任务需求等时刻重新规划队形,保持编队队形的完整性。通过仿真与物理实验验证了基于竞标机制的时间最优实时编队规划算法的可行性,与其他编队规划算法相比,所提出的编队时间最优策略在编队形成中具有时间短、效率高的优点。3)利用基本行为服务与监督机制,研究多移动机器人编队形成和运动过程中出现障碍物时的避障策略。本文提出了基于多动态虚拟目标跟踪的编队避障策略,同时考虑了编队和障碍物信息。将具有多动态虚拟目标的跟踪方法与基本运动行为服务结合,实现编队运动中移动机器人能脱离U型槽障碍物,继续形成编队运动,同时采用监督机制保证编队机器人的避障过程中编队队形不会发生改变。通过多机器人编队运动的避障实验,表明本文提出的基于多动态虚拟目标跟踪的编队避障策略的有效性。4)利用基本行为服务和避障策略,在编队控制的基础上,研究多移动机器人在开阔平坦环境中的协作运输问题。与已有的反应式行为协作运输不同,本文以观测者—推动者的协作关系,将运输任务转换成多移动机器人编队问题,考虑了实际机器人的传感器和搬运机构的限制条件。为减少反应式行为控制方法所导致的动作重复执行的缺陷,利用几何原理对编队运输的运动和避障路径进行规划,通过设置虚拟目标点,对虚拟目标进行跟踪,完成编队运输的任务。在Microsoft Robotics Developer Studio (MRDS)仿真环境中,设计推箱子实验并与反应式行为控制方法比较,验证协作运输策略的可行性。5)研究了起伏地形环境下的多移动机器人协作运输问题。起伏地形环境中由于地形限制,机器人的侧滑、驱动轮形变、摩擦和环境难以区分等,使得多机器人协作运输面临更多困难。本文提出了Navigate Straight Forward to Destination (NSFD)和Navigate Plain First to Destination (NPFD)导航策略,设计了推动者机器人发生滑动时编队重新形成策略。在起伏地形的仿真环境中,设计了协作推箱子实验,验证了协作运输策略的有效性,并对NSFD和NPFD两种导航策略在不同环境下的协作运输效果进行了比较,结果表明NPFD更适合于复杂地形环境下的协作运输。6)以多移动机器人编队与单任务为基础,研究多任务机器人—多机器人任务(MT-MR)的任务分配问题,研究多个任务并存时的机器人联盟建立问题。本文考虑机器人能源有限及携带传感器不同的情况,提出了一种具有次优解的基于信用机制的动态任务联盟形成方法CoSGCrM。通过与first-price竞拍方法进行比较,验证了提出的多机器人联盟多任务分配方法的可行性,并具有较高的任务完成效率。更多还原

【Abstract】 Multiple mobile robots system provides an unchallenged incentive to all researchers in the past two decades, with its distribution characters in time and space, as well as its high efficiency. Great interests have been attracted to application domains including industry, military, national defense, daily life, and deep space, etc. In the research of specific application tasks, such as transporting heavy object, arresting target and multi-sensors cooperative map exploration, etc., robots are required to form and maintain specific shapes of formation during the tasks’execution. And these lead to a rush in the research of multi-robot formation control, which is one of the key techniques in multi-robot cooperation and coordination. Takes the heavy object transport problem as application background, problems in formation control are discussed in this thesis, from low level behavior-based motion control, motion planning, to high level formation task planning.Contributions of this thesis can be summarized in several points illustrated below:1. As the foundation of robot movement and formation control, the basic robot motion behavior and the behavior structure are designed at first. In the existing works on behavior-based control, behaviors are typically designed with focus on specific tasks or experiments, most of which are high level descriptive behaviors. The design of basic behavior and its reusability are less mentioned. In this thesis, the behaviors of multi-robot formation are proposed, with the posture deviation between robot and virtual target robot in formation as its input, and the driven wheel rotation velocity as its output. Service-oriented architecture (SOA) with reusability is introduced to the design of the basic motion behavior structure. This structure can satisfy new conditions with proper modification, extension or combination. The feasibility and reusability of the proposed service-based motion behavior structure are proved in both simulation and experiment. The structure is used in researches on multi-robot formation control and cooperative transport tasks.2. On the basis of basic motion behavior, formation initialization and movement planning are studied. An auction based formation initialization planning method with time optimal index is proposed. The formation problem has been transformed to task allocation problem, that is, the mobile robots in follower roles bid for the position where the virtual target robots is in formation, and the robot in leader role performs a centralized planning algorithm to allocate the positions to all followers. The integration of the centralized planning and distributed biddings can accomplish the position allocation task while reducing the complex of computation dimension. Besides, the real-time monitor mechanism raised at the bidding moment can help the leader to plan the formation shape again as soon as robots break down, or task requires. This mechanism can maintain the completeness of the formation. Simulation and experiments in formation control are taken to illustrate the feasibility of the proposed auction-based time optimal strategy. The formation time-optimal strategy is compared with single robot time shortest strategy and the strategy that robot and target with the same suffix are preferred. Results show that the proposed strategy cost less time and has higher efficiency in formation generation.3. Obstacle avoidance strategy in formation generation and movement is studied with the proposed basic motion behavior service and monitor mechanism. A multiple dynamic virtual targets tracking based obstacle avoidance strategy is proposed, takes both the formation information and the obstacle information into consideration. The proposed obstacle avoidance strategy combined with the motion behavior can accomplish the task that the robot in formation escapes from U-shape trough and returns to the position in formation. Monitor mechanism ensures the formation shape would be kept while robot is avoiding obstacle. Experiment of obstacle avoidance in formation illustrates the effectiveness of the proposed multiple dynamic virtual targets tracking based obstacle avoidance strategy.4. Multi-robot cooperative object transport task is studied on the basis of formation control, with the basic motion behavior service and obstacle avoidance strategy. Being different from the existing reactive behavior based cooperative transport method, the cooperative transport problem with robots in the relationship of pusher-watcher is transformed into the multiple mobile robots formation control problem. Limits on real robot sensors and actuators are also considered. In order to reduce the defects of the repetition movement controlled by reactive behaviors, geometric relation is considered in the path planning of the movement and obstacle avoidance of the formation. The virtual target in the path is set for robot formation to track, so robots can accomplish the transport task. Box pushing task with compare to reactive behavior-based approach in Microsoft Robotics Developer Studio (MRDS) is designed to illustrate the feasibility of the proposed cooperative transport strategy.5. Multiple mobile robot transport problem in the environment with undulating terrain is studied. With the restraint in terrain, as well as slipping of the robot, deformation of wheel, friction etc., it is more difficult for robots transport task. Navigation strategies of both Navigate Straight Forward to Destination (NSFD) and Navigate Plain First to Destination (NPFD) are proposed in this thesis. Formation rearrangement strategy is designed when pusher robots slipped. Simulation of the box pushing in the environment with undulating terrain is performed to illustrate the validity of the proposed strategy. The cooperative transport strategies with NSFD and NPFD are compared in different environment. The results show that the one with NPFD is more suitable for cooperative transport task in the complex environment.6. On the basis of the multiple mobile robot formation control and the single task allocation problem, task allocation problem of multi-task robots-multi-robot tasks (MT-MR) is studied. Considering the energy resource limits and the different sensors equipped in robot, a dynamic coalition structure generation based on credit mechanism (CoSGCrM) approach is proposed with sub-optimal solutions to the coalition formation problem. Compared with the first-price auction method, the feasibility and the higher efficiency in accomplishing tasks of the proposed multi-robot coalition formation task allocation approach has been proved.更多还原