弱观测复杂海洋环境下AUV动态目标跟踪算法研究

Compared to the large underwater base station, there is an inherent lack of aperture in the passive flank array sonar installed on the autonomous underwater vehicle (AUV). And therefore a series of scientific problems were raised, such as limited range and inaccurate bearing angle measurements. In order to estimate motion parameters of the dynamic target fast and accurately in an environment subjected to multiple noisy sources, low signal-to-noise ratio and weak observation, research mainly focus on the following problems: 1) The timely stable target motion analysis filter algorithm which is robust to varying system noise and measurement noise over time; 2) the tracking gate and data association algorithm used in the adverse environment subjected to multiple noisy sources and low signal-to-noise ratio; 3) the tracking strategy capable of meeting the needs of estimate performance and the shortcut constraints for the weak maneuvering AUV; 4) the design and implementation of a demonstration simulation system for the mission scenario of AUV's passive tracking and the validation for the core tracking strategy on the simulation system The adaptive double unscented Kalman filter, flexible tracking gate, joint probabilistic data association algorithm base on artificial neural network and OTP algorithm based on inclusive differentiation are adopted and refined on. With the help of mathematic modeling, theoretical analysis, simulation experiment and field data validation, the rules of inherent spatiotemporal coupling between motion parameters estimating performance and the momentarily changing situation formed by AUV and dynamic target are studied. This project will make an important contribution to forming the basis of further behavior decision-making of AUV. Furthermore, it is also of importance signification for laying a solid theoretical foundation for the research and development of the AUV with higher autonomy.

与大型水下基站相比，自主水下机器人(AUV)搭载的舷侧阵声纳存在固有孔径不足问题。针对该被动声纳受限引起的一系列科学问题，以提高多杂波、低信噪比、弱观测复杂海洋环境下目标运动参数估计的快速准确性为目标，主要研究：1)弱观测条件下系统噪声自适应的快速稳定的TMA滤波算法；2)多杂波、低信噪比复杂海洋环境下的柔性跟踪门及数据关联算法；3)同时满足参数估计性能和弱机动AUV捷径跟踪等约束的轨迹规划算法；4)AUV对水下动态目标被动跟踪仿真系统及核心算法验证。采用基于自适应双无色卡尔曼滤波算法、柔性跟踪门、神经网络联合概率数据关联算法和基于微分包含的OTP算法，探索复杂环境下AUV对水下目标运动参数估计性能与观测平台机动轨迹之间内在的时空耦合规律，并通过数学建模、理论分析、仿真和现场数据验证等手段进行研究。本项目研究成果将为AUV进一步的行为决策提供依据，为更自主AUV的研发奠定坚实的理论基础。

对目标的探测、定位、跟踪、决策及控制是无人系统完成其使命任务的关键步骤，特别对非合作目标探测常采用被动方式。鉴于自主水下机器人具有机动灵活能执行多种水下使命任务等优势，本项目以AUV为观测平台被动跟踪水下动态目标。在复杂的海洋环境下，AUV受到水下环境噪声、平台噪声、多杂波、低信噪比、弱机动性等因素限制。同时，由于尺度限制无法搭载大尺度探测传感器，由此引起了一系列难题。针对上述问题，深入研究以下内容。.(1)根据AUV被动目标跟踪特点，选择合理的坐标系和目标运动模型，建立AUV被动目标跟踪数学模型，根据被动探测声纳的传输模型、目标噪声模型及其AUV探测传感器模型，建立用于仿真平台的被动声纳目标探测模型。.(2)针对传感器受限引起的初始距离误差较大和复杂的海洋环境下系统噪声统计不精确问题，利用强跟踪滤波器原理建立基于强跟踪的SRCKF滤波算法，仿真实验表明该方法具有较低的关于系统噪声及初始状态统计特性的敏感性等优点。针对模型不确定性引起的鲁棒性差问题，推导出了鲁棒滤波算法，并在修正极坐标系和笛卡尔坐标系下进行了仿真验证。针对机动目标跟踪问题，提出了一种基于实时辨识自适应滤波的改进交互式多模型算法，根据目标的机动强度对过程噪声方程进行自适应调整，通过仿真验证算法的有效性。.(3)针对密集多杂波环境下的数据关联问题，分别采用贝叶斯框架下的基于无味变换的最大熵模糊概率数据关联滤波器、联合概率数据关联滤波器、广义概率数据关联滤波器和基于粒子滤波的数据关联滤波器，实现水下单AUV和双AUV在多杂波环境下的单目标及多目标被动跟踪，通过数值仿真实验和外场数据实验对上述方法进行了验证和分析。.(4)针对AUV机动能力和传感器探测范围受限等约束问题，提出了基于多约束的AUV轨迹优化方法，并通过仿真验证了该方法的可行性和有效性。针对复杂海洋环境不确定性问题，提出了基于策略专家系统的AUV航路规划方法，给出了策略专家系统工作原理，通过半物理仿真和外场试验验证了系统的可行性。