【作者】 周彤；

【导师】 洪炳镕；

【作者基本信息】 哈尔滨工业大学 ， 计算机应用技术， 2008， 博士

【摘要】 由于机器人成本的降低和性能的提高,使用移动机器人作为传感器网络的节点已经可行,这种移动传感器网络形式具有无线分布式传感器网络所有的功能,而且节点具有可移动性,扩大了传感器网络的覆盖范围,提高了传感器网络的感知能力,它可以广泛应用于外星球探索、军事侦察、灾难搜救等领域。多移动机器人具有并行处理、容错、柔性和信息冗余的优点,使用多机器人的协作功能扩展了单个机器人的能力。本文在国家863计划资助项目“分布式多机器人合作与竞争机制及其应用技术”和国家自然科学基金资助项目“基于不精确地图的移动机器人室内导航技术”的资助下,以拓展移动机器人应用领域为目的,以理论研究和仿真实验为手段对移动传感器网络需要的关键技术进行了研究,提出了移动传感器网络节点在覆盖、部署、自定位以及目标跟踪方面的一些新算法。首先介绍了多机器人系统的特点和国内外多机器人系统研究方面的现状,并简要介绍了多机器人协作的概念、分类、需要解决的关键问题等,进而引出由多机器人系统组成的移动传感器网络概念、体系结构、拓扑结构的种类和系统评价标准,并介绍了移动传感器网络的应用领域,研究现状,以及需要解决的关键问题。由于移动传感器网络是由移动机器人构成,具有移动的特性,可以扩大感知区域的覆盖范围。本文对移动传感器网络中机器人节点常使用的动态覆盖算法进行了介绍,由于通常的动态完全覆盖算法主要使用在规则的感知区域,针对移动传感器网络实际应用中的感知区域的不规则性的限制,提出了一种基于覆盖误差控制节点移动的算法,这种算法基于对覆盖误差梯度的计算,控制每个移动机器人节点移动速度和移动方向,同时还考虑了多节点移动的避碰问题。本文不仅在理论上进行了控制规则的分析,同时用仿真的方式进行了算法验证,此算法可以在灾难救援等实际应用中使用。节点部署是移动传感器网络研究领域另一个重要课题,本文介绍了几种经常用于移动传感器网络节点部署的算法,同时提出了基于虚拟力的传感器网络节点部署算法,这种算法不仅使用传感器范围内的节点之间的距离信息,而且还考虑了节点间进行通讯获得的邻居节点间的位置信息,通过这种分布式控制节点移动方式,达到在待感知区域均匀部署节点的目的,并最大化覆盖待感知区域。这种控制算法不需预知环境地图信息,具有分布式控制效果,且具有自适应性,可以应用于军事侦察领域。由于感知对象的信息和其物理位置信息密不可分,所以节点自定位是传感器网络应用中传感器节点必须具备的功能,本文介绍了基于测距和基于非测距的节点自定位算法,并根据蒙特卡罗法,提出了一种基于混合跳跃蒙特卡罗法HMCL的节点自定位算法,此算法可以根据锚节点数量的多少,自行选择适合的定位策略,计算自己的位置坐标,通过与其它近似算法的比较,此算法具有自适应性和较高的准确性。由多机器人组成的移动传感器网络的一个最广泛应用领域就是对动态目标的观测及跟踪。本文基于人工势场法,提出了基于局部机器人及目标密度的多目标跟踪算法LRTDA,根据机器人节点密度和目标密度,进行虚拟力控制参数的调整,达到分布式跟踪多个移动目标的目的,提高了对移动目标的观测率和传感器网络的覆盖率,仿真实验表明了这种算法的有效性和可行性。此算法可以应用于战场侦察和灾难救援等领域。更多还原

【Abstract】 As the cost reduced and the performance improved, using autonomous mobile robots as nodes of sensor network has become feasible. This kind of mobile sensor network has all functions that wireless distributed sensor network has. Its mobility increases the coverage range and upgrades the sensing ability of sensor network. Mobile sensor network can be extensively used in planetary exploration, military reconnaissance and disaster rescue. Multi-robots have such advantages as parallelism, flexibility, fault tolerance and data redundancy. Multi-robot coordination can extend the capability of the single robot.Supported by the national high-tech research and development plan of China,“Collaboration and Competition Mechanism for Distributed Multi-robots and its Application Techniques”and the NSF project,“Indoor Navigation Techniques of Mobile Robot based on Imprecise Map”, this dissertation aims to exploiting on the application domains of the mobile robotics and systemically studies several key technologies existing in mobile sensor network by theroretical and simulated way. Several novel algorithms are proposed for mobile sensor nodes including coverage, deployment, self-localization and multi-target tracking.The features and present research for multi-robot system are introduced firstly. The concept, classification and existing problems for multi-robot coordination are briefly presented. The concepts, system architecture, species of topology structure and evaluation criterions of system for mobile sensor network composed of multi-robot are elicited. And the application domains, present research state and key problems required to resolve are introduced.Mobile sensor networks are constructed by mobile robots. The robots have the mobility to increase the coverage for sensed region. Several common dynamic coverage algorithms are introduced firstly. These algorithms are used in regulation regions. An algorithm based on coverage error is proposed to control the motion of mobile nodes to overcome the limitations of abnormity for sensed regions. This algorithm is based on the computation for the grads of coverage error to control the speed and direction for robots and the avoid collision between nodes are considered at the same time. The control rules are analysed by theoretically and verified by the simulations. The algorithm can be used in disaster rescue.Nodes deployment is an importmant subject for mobile sensor network domain. This paper introduces several common deployment algorithms for mobile sensor networks. And the mobile sensor nodes deployment is proposed based on local virtual force. The algorithm utilizes not only the distance informations among nodes but also the position informations through communication between nodes. This distributed control way for nodes moving to achieve the goal of nodes uniformly deployment and maximal coverage in the sensed area. This algorithm need not know the map information of environment previously. It can be used for battle field due to the distributed control effectness and self-adaptability.Sensor nodes are required to have the function of self-localization because the sensed targets usually associate with geographical information in sensor network. Self-localization algorithms based on range and range-free are introduced firstly. This paper presents a hybrid hops Monte Carlo self-localization algorithm (HMCL) based on Monte Carlo Localization. The algorithm may select proper localization strategy to compute the coordinates of nodes according to the numbers of anchor nodes in sensor network. The algorithm is self-adaptable and higher accuracy compared with other similar algorithms.The most extensive application domain is to observe and track dynamic targets by mobile sensor network composed of multi-robot. The Local Robot Target Density Alogrithm (LRTDA) is proposed based on the Artificial Potential Fields. Based on the local density of robots and targets, the control parameter of virtual force is adjusted for achieving to track multi-target. The algorithm improves the observation for multi-target and coverage for sensor network. Simulation shows the validity and feasibility for this algorithm. This algorithm can be used for these domains, such as reconnaissance in battle field and rescue in disaster area.更多还原