【作者】 邹启杰；

【导师】 张汝波；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2014， 博士

【摘要】 水面无人艇(Unmanned Surface Vehicle,简称USV),是一种具有一定的自主能力的、无人操作的水面机器人系统。通过挂载智能任务控制系统,USV能够在远离用户的海洋环境下,独立执行危险以及其他不适于有人船只执行的任务。正是具有这样的任务执行能力,USV得到了各国海军的广泛重视。近些年,随着USV的研究和发展,其自主能力得到不断的提升,同时人们对于USV执行复杂任务的期望也越来越高,引发不确定事件处理能力有限的问题。对于此类问题,有研究人员提出利用可变自主机制来解决。但是,对于可变自主实现机制的研究目前仍处于起步阶段,存在诸多挑战性问题。因此,对于USV可变自主实现机制的探索,无论是从理论意义还是应用前景来说,都具有重要的研究价值。本文将围绕着USV可变自主实现机制为核心,分别研究USV可变自主系统体系结构模型、上下文感知的态势推理、用户因素建模机制以及自主等级决策方法。本文的主要研究工作概括如下:(1)研究了一种支持可变自主控制的USV体系结构模型,解决了针对多种不确定性因素影响下用户动态加入USV控制过程的问题。将传统观察、判断、决策和行动的控制流程与多目标决策理论相结合,建立OOADA(Observe-Orient-Adjustable Autonomy-Decide-Act,OOADA)的可变自主控制流程。进一步,研究了具体的USV可变自主系统体系结构模型,其中包括关键性的可变自主控制模块的详细构造。通过USV仿真系统的搭建以及导航任务的实验,验证了该模型能够有效解决USV自主等级决策、控制与实施的问题。(2)以导航任务为重点,研究了基于上下文感知理论的态势推理方法。针对USV自主能力受多种不确定性影响的问题,根据上下文感知的理论基础,研究了人机交互的推理方法。该方法包括表征USV和用户交互的各种不确定信息的采集,以及利用期望效用函数进行USV系统性能分析。在USV执行导航任务的仿真实验过程中,模拟多种不确定性事件以验证USV关于人机交互的推理能力。实验结果证明,利用了上下感知推理,能够帮助USV有效识别交互需求,准确预测不同自主等级下的交互结果,为自主等级决策提供全面的数据依据。(3)研究了 USV可变自主系统中的用户模型,将用户因素分两个方面与自主等级决策建立关联。其中,用户偏好是利用多属性决策理论评价用户工作负荷对于自主等级的影响;用户态势认知能力是通过马尔可夫决策模型预测用户认知结果对于自主等级的影响。通过仿真实验,利用USV导航任务分别测试同一用户不同状态和不同用户对于自主等级决策的影响。验证结果说明USV区分不同的用户认知能力是必要的,同时用户模型在系统中有效控制了用户的工作负荷。(4)研究了基于多属性决策的USV自主等级决策方法。引入组织排序法解决USV自主等决策目标不可公度的问题,利用优先函数评价不同等级之间每个性能目标的优势,减少了数据预处理带来的评价误差;利用赋值有向图对每个自主等级进行效用并合,建立全部自主等级的完全序,获得自主等级决策的最优解。仿真实验对比了可变自主机制与全自主机制下USV导航能力的差异,表明可变自主对于系统性能的改善十分明显。此外,全面分析USV系统的整体性能,验证USV可变自主系统以较少的用户参与,获得较大程度的系统性能的提升。更多还原

【Abstract】 Unmanned surface vehicle,referred as USV,is a kind of surface vessels having certain autonomy and can be operated on the water surface without human.By loading intelligent task system,USV can carry out dangerous tasks and tasks unsuitable for boat with man,including different kinds of military tasks and non-military tasks.Thus US Vs are getting much attention by navies of different countries.These years,along with the development of the USV,its autonomy ability has been increased greatly.Meanwhile people’s expectations for USV performing complex tasks increase highly too.The problem of limited ability of USV on dealing with uncertain events thus emerges.Currently researchers agree that the use of adjustable autonomy mechanism can solve the problem.But there are still many challenges for USVs surfing on the sea surface.Therefore,explore on adjustable autonomy mechanisms of USVs has important research value,whether from a theoretical or application prospect.This thesis takes the mechanism of adjustable autonomy for USVs as the core problem for research,and researches the structure of adjustable autonomy system model for USVs,the decision method for autonomy levels,the situation awareness based on context awareness and the modeling mechanism of users’ factor respectively.The main research work of this paper can be summarized as followings:(1)A model of system supporting adjustable autonomy control mode is researched,which solves the problem of how user joins the control dynamically while influenced by variety of uncertain factors.The traditional control process of observation,orientation,decision-making and action is integrated with multi-objective decision theory,thus build up new control process of Observe-Orient-Adjustable Autonomy-Decide-Act,OOADA.Further,the adjustable autonomy system model for USVs is researched in detail,including the detailed design of adjustable autonomy module.Proved by deployed in simulation platform and experiments,the adjustable autonomy system model based multi-object decision theory can realize adjustable autonomy control mode.And because the structure of this model is build upon decision theory,it has solid theoretical foundation and can support the expansion requirements of task,object sets and diversity of autonomy levels,which have great and general value.(2)For navigation tasks,the method for situation awareness the contextual perception model in the adaptive autonomy control mode is researched,including context information collecting and reasoning.For the problem of autonomy ability of USVs often intervening by a variety uncertainties,the method for reasoning situation of human-USV interacting is build based on context awareness theory.The method includes definition and collection of the human-USV interaction information and the USV system performance analysis used multi-attribute utility function.Simulation experiments simulate different kinds of uncertain events to testify the ability of USV’s situation reasoning during the navigation tasks.The result shows uncertain events monitoring can identify different types and different degrees of uncertain events.At the same time it can also prove that context reasoning model can predict the different performance indicators’ trends under different autonomy levels,which can satisfy the needs of autonomy decision on system perform reasoning before and after the adjustment.(3)The thesis researches the user model under the contextual awareness by dividing it into two dimensions:the user preferences and user situation cognitive ability,which are both used to connect with autonomy levels.The user performance uses multi-attribute decision theory to evaluate the influence of users’ workload on the autonomy level.While user situation cognitive ability takes Markov decision model predict the influence of the result of user cognition on autonomy levels.In the simulation experiments,USV navigation tasks are used to test the situation recognition ability of the system on different users and same user with different preferences,along with the impact on the result of autonomy levels decision.The results prove that identifying user is necessary and beneficial.At the same time the user model controls the workload of the user,and decreases the instability of the system performance brought by changing users.(4)The paper researches the algorithm based on multiple attribute decision making to decide the autonomy level of USV.Using preference ranking organization method for enrichment evaluation solve the problem such as incommensurability of autonomy decision objects.The priority functions are defined to evaluate the advantages of each performance indicator,thus avoid the estimation error brought by data preprocessing.Then the algorithm builds a directed diagraph to merge the utility of each autonomy level,after which the difference advantages of autonomy levels are get.Simulation experiments of comparing fully autonomy with adjustable autonomy based on multi-object decision theory are conducted to show the changing of the system performance in navigation task.The performance of the USV system under adjustable autonomy mode is analyzed and evaluated comprehensively.The results prove that the adjustable autonomy based multi-object decision can dramatically increased system performance with a little increasing of user participation.更多还原