面向机动目标跟踪的多不确定参数耦合系统状态估计与融合

The state estimation and fusion for non-cooperative maneuvering target tracking with multiple sensors are challenged by multiple uncertain parameters coupling characteristics, such as multi-mode, nonlinearity, noises correlation, additive and multiplicative noises coexistence, distributed architecture, unknown disturbance and gaming. In this proposal, on the basis of modeling the multi-mode uncertainty obeying a Markov chain, the highly accurate and fast filter with distributed fusion strategy is designed for the related multiple uncertain parameters coupling dynamical systems. Considering the coupling coexistence of the multi-mode uncertainty and randomness of system matrices caused by the unknown relationship between maneuvering targets and echoes, a general filter framework about state estimation and data association in clutters is established for tracking maneuvering targets, and a distributed fusion strategy is further proposed in the multi-sensor condition. Facing with colored measurement noises, additive and multiplicative noises, multi-mode uncertainty and model nonlinearity deeply coupling in complex environment, an adaptive Gaussian sum filter is designed and a distributed fusion method is then presented via consensus strategy with decentralized multiple sensors. In a target chasing involving two maneuvering platforms perturbed by unknown disturbances, both platforms desire to have a precise estimate about the other side and avoid to be caught up. By utilizing the mutual information to measure the uncertainty coming from the measurement sequence, an optimizing platform moving strategy is discussed and the condition about the related system approximating dynamic equilibrium is analyzed. This proposal studies the joint optimization about state estimation and related uncertainties in the global sense, in order to improve the quality of maneuvering target tracking, and also extends the state estimation theory in complex environment.

面向多传感器非合作机动目标跟踪的状态估计与融合越来越呈现出多模态、非线性、噪声相关、加性/乘性噪声共存、分布式结构、干扰博弈等诸多不确定参数耦合特性。本项目拟基于跳变马尔可夫过程刻画多模不确定，设计相应不确定参数耦合系统的高精度快速滤波及分布式融合策略。在多传感器杂波环境下，考虑多模不确定与目标/回波对应关系不确定引起系统矩阵随机耦合共存，建立状态估计与数据关联一体化滤波框架并设计分布式融合策略。针对复杂环境下量测噪声有色、加性/乘性噪声共存、多模不确定与模型非线性等深度耦合，设计自适应高斯和滤波及分布式融合共识策略。考虑未知扰动下机动双平台博弈跟踪，双方皆希望准确估计对方且有效躲避对方跟踪，借助互信息度量量测序列不确定性大小，设计平台移动优化策略并分析系统趋于动态平衡的条件。项目研究在全局意义下实现状态估计与相关不确定性的联合处理，以便改善目标跟踪质量，并扩展复杂环境下的状态估计理论。

多传感器非合作机动目标跟踪的状态估计与融合总是面临探测/传输特性（诸如加性/乘性噪声共存、分布式架构、博弈对抗、随机时滞等）、目标自身衍化特性（诸如模型非线性、多模态等）和复杂外界环境（诸如噪声有色、噪声相关、噪声非高斯等）相互影响而引起的多不确定性耦合共存现象。为有效提高典型场景下机动/弱机动目标跟踪质量，实现全局意义下状态估计与多不确定性的联合处理，本项目主要开展了下述研究工作：以随机系数矩阵（乘性随机参数）刻画机动目标-回波间关联不确定性，推导了面向杂波环境下噪声相关的机动多目标跟踪递推线性最优滤波及分布式实现，构建了状态估计与数据关联一体化滤波框架，并设计了区间约束下递推上限滤波；基于高斯/高斯混合分布近似后验概率密度，提出了相关噪声/非高斯噪声环境下面向机动/弱机动目标跟踪的自适应高斯混合滤波及分布式融合估计，并探索了基于多重重要性采样的确定性混合粒子滤波；借助互信息度量未来量测包含的目标有效信息大小，构建了敌-我双方单平台对等机动能力下的博弈跟踪模型，统计、归纳、分析了典型运动轨迹模式与帕累托优化参数取值间的关系；为克服WiFi信号下目标信道状态信息相位随时间变化不可预测，合作探索了迁移学习下设备无关被动定位算法。上述研究工作的开展，丰富和发展了目标跟踪与多传感器融合的新理论和新方法，初步为现代战场复杂环境下可靠、精确、快速的多机群自主跟踪博弈国防前沿探索了新思想和新途径。