【作者】 丁力；

【导师】 刘梅；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2009， 硕士

【摘要】 随着运载火箭和航天器技术的飞速发展,人类的活动空间从陆地、海洋、天空扩展到了太空。在新军事变革的引领下,太空成为未来战争的制高点,制太空信息权是夺取高科技战争胜利的必备因素。卫星网络作为实现太空信息对抗的有效手段,具有重要的军事价值和广阔的发展前景。然而,卫星网络节点数量众多,单个节点的感知、处理能力有限,网络拓扑结构动态变化,应用环境复杂多样,这些特点从基础理论和工程实际两个层面给网络的协同应用带来巨大挑战。因此,本文围绕大规模高动态立体卫星网络协同技术这个核心,以跟踪多个空间机动目标为背景,首次将多智能体理论引入到具有高动态拓扑结构的空间网络中,构建了基于移动动态联盟的卫星网络多机动目标协同跟踪理论和仿真模型,并深入研究了其中的几项关键技术。本文主要研究内容概述如下:1.针对空间移动目标跟踪这个背景,结合移动智能体和动态联盟的基本思想,引入移动动态联盟的概念,并以此为基础发展了基于移动动态联盟的卫星网络多机动目标协同跟踪理论。该理论将协同跟踪过程分为动态联盟形成、联盟内部的协同信息处理和动态联盟交接三个步骤,从理论上解决了大规模高动态立体卫星网络协同跟踪多个空间移动目标的难题。2.针对卫星网络协同跟踪多目标时的动态联盟形成问题,提出了一种基于合作博弈的动态联盟形成算法。首先采用弹性神经网络模型对卫星网络进行多任务集合划分,然后对多动态联盟形成问题建模,提出了基于合作博弈的移动动态联盟形成算法,解决了多目标跟踪时多个动态联盟自组织形成的难题,同时在整体的跟踪过程中提高了卫星网络的跟踪精度。3.针对协同信息处理技术中的数据融合问题,将非线性滤波算法和平均一致技术相结合,提出了完全分布式的数据融合算法,解决了原有数据融合技术应用在卫星网络中鲁棒性差、生存能力不强的缺陷;然后对标准平均一致算法进行改进,提高了算法的收敛速度。4.针对协同信息处理技术中的融合结构优化问题,将Gossip路由协议和完全分布式的融合算法相结合,并利用次梯度算法进一步改进信息传输结构,解决了信息广播带来的内爆问题。5.针对动态联盟移动过程中的交接问题,构造了由预测机制和修复机制相结合的动态联盟交接方法。此方法不仅设置了预测机制来更新下一时刻动态联盟的跟踪覆盖区域,而且针对目标出现大机动这种特殊情况,引入了基于内插的修复机制,提高了机动目标的跟踪精度,解决了卫星网络对高机动目标协同跟踪并建立连续航迹的难题。6.针对卫星网络协同跟踪多机动目标的实际应用,建立了完整的移动动态联盟多机动目标协同跟踪模型。仿真实验验证了该协同跟踪模型的有效性。更多还原

【Abstract】 With the rapid development of launch vehicle and spacecraft technology, human activities extend to the outer space from the land, sea and space. Under the guidance of the new military revolution, outer space has become the dominant position of the future wars and information dominance in outer space is an essential factor to win the high-tech wars. As an effective means to carrying out information countermeasures in the outer space, satellite network has an important military value and broad prospects for development.However, the number of nodes in satellite network is large; the detection and processing capability of single node is limited; the topological structure of satellite network dynamically changes; the application environment of satellite network is complex and diverse. These characteristics bring on tremendous challenges in collaborative applications of satellite network from basic theory to engineering application. Therefore, aiming at collaborative technique of large-scale, greatly dynamic and tridimensional satellite network, under the background of tracking multiple maneuvering targets in outer space, this paper applies multi-agent theory to a greatly dynamic network in outer space firstly, then constructs integrated multi-target collaborative tracking theory and simulation model based on mobile dynamic coalition, moreover studies several key techniques. The following is the summarization of the main work in this paper:1. Aiming at the background of tracking moving targets in space, the concept of mobile dynamic coalition is presented combining the basic idea of mobile agent and dynamic coalition. Then integrated multi-target collaborative tracking theory is developed based on mobile dynamic coalition in satellite network. The theory divides collaborative tracking process into several steps such as the dynamic coalition formation, collaborative signal and information processing in dynamic coalition and dynamic coalition handover. It solves the difficult problem of a large-scale, greatly dynamic and tridimensional satellite network collaborative tracking multiple maneuvering targets in theory.2. Aiming at the dynamic coalition formation of collaborative tracking multiple maneuvering targets in satellite network, a coalition formation method is presented based on cooperative game. Elastic neural network model is utilized to partition satellite network into several node sets firstly. Then the model of multiple dynamic coalitions formation problem is established and the mobile dynamic coalition formation algorithm is proposed based on the cooperative game. The proposed algorithm solves the difficult problem of multiple dynamic coalitions self-organizing formation, moreover improves tracking accuracy of satellite network in the whole tracking process.3. Aiming at the problem of data fusion in collaborative signal and information processing technique, a completely distributed data fusion algorithm is presented combining nonlinear filter algorithm and average consensus technique. The proposed algorithm solves the disadvantages such as weak robustness and viability when old data fusion technique applies in satellite network. Then an improved method is proposed to accelerate average consensus algorithm achieving convergence.4. Aiming at the problem of fusion structure optimization in collaborative signal and information processing technique, an algorithm is proposed combining Gossip routing protocol and completely distributed data fusion technique. Then subgradient method is utilized to improve information transmission structure. The proposed algorithm solves implosion problem resulting from information broadcasted.5. Aiming at the problem of handover in the dynamic coalition moving process, a dynamic coalition handover method comprising of the predicting mechanism and the recovery mechanism is constructed. This method not only sets up a predicting mechanism to update the tracking coverage region of the new dynamic coalition, but also aiming at the special situations that targets maneuver, presents a recovery mechanism based on interpolation. This handover method improves tracking accuracy of maneuvering targets and solves the difficult problem of satellite network tracking highly maneuvering targets and setting up continuous tracks.6. Aiming at the practical application of satellite network collaborative tracking multiple maneuvering targets, integrated multi-target collaborative tracking simulation model is constructed based on mobile dynamic coalition. Simulation results validate the effectiveness of this collaborative tracking model.更多还原