近年来随着信息技术的迅猛发展，深度融合了计算、通信与控制能力的信息物理系统成为当下热门的交叉研究领域。状态估计系统是保证信息物理系统稳定高效运行的基础与关键所在。然而由于传输网络的信道带宽有限且时常伴随丢包、网络攻击等问题，信息物理系统的状态估计系统面临巨大挑战。本文针对信息物理系统远端状态估计系统面临的传输网络信道带宽有限、数据丢包、网络攻击等问题，系统地开展复杂信道条件下基于事件驱动的容积Kalman滤波算法研究。 针对信息物理系统传输网络信道带宽有限且对系统状态估计精度及实时性要求较高的问题，开展了事件驱动容积Kalman滤波器设计研究。利用基于新息条件的事件驱动采样策略减少数据发送量以节省带宽；利用滤波精度高、数值稳定性强且计算量相对较少的容积Kalman滤波器设计了事件驱动非线性滤波器以保证滤波精度；对设计的事件驱动容积Kalman滤波器开展了随机稳定性的分析研究，利用随机Lyapunov方法证明了该滤波器随机稳定性，给出了完全由离线参数组成的滤波器随机稳定充分条件，推导了新息条件下事件驱动采样策略的平均通信到达率，为实际工程中事件驱动容积Kalman滤波器的设计提供了理论参考。 针对信息物理系统的系统噪声未知且量测噪声为非高斯情况下的状态估计问题，开展了事件驱动鲁棒容积Kalman滤波算法研究。针对新息条件事件驱动采样策略的不足，利用随机新息条件作为事件驱动采样策略保证量测新息统计特性服从高斯分布，并设计了随机新息条件下事件驱动容积Kalman滤波器；在此基础上分别利用滑窗估计方法与自适应方法对量测噪声及过程噪声进行估计，并利用Huber方程对随机新息条件下事件驱动容积Kalman滤波器进行鲁棒性设计以适应未知及非高斯噪声；分析了上述两种滤波器的随机稳定性，给出了滤波器随机稳定的充分条件。 针对信息物理系统传输网络存在信道丢包情况下状态估计问题，开展了事件驱动容积次优Kalman滤波器设计研究。针对以周期采样作为采样策略时通信网络信道存在丢包的问题，以独立同分布Bernoulli随机过程对信道丢包进行建模，设计了次优容积Kalman滤波器；针对事件驱动采样策略下信道存在丢包问题，利用随机新息条件设计了事件驱动采样策略，并结合次优容积Kalman滤波器设计了事件驱动次优容积Kalman滤波器；利用随机Lyapunov方法证明了上述两滤波器的随机稳定性，并给出了滤波器随机稳定充分条件。 针对信息物理系统遭受网络攻击情况下的状态估计问题，开展了网络攻击情况下事件驱动滤波器设计研究。对数据篡改和偏差型控制命令伪造攻击两种常见攻击方式进行了建模；针对数据篡改攻击问题，利用投影统计方法设计了异常数据检测器，根据检测结果构建了权重矩阵以矫正量测值保证滤波精度；针对偏差型控制命令伪造攻击问题，将其转化为系统有未知输入问题并利用贝叶斯推理方法详细推导设计了此类攻击下事件驱动容积Kalman滤波迭代算法。 关键词：信息物理系统；事件驱动采样策略；容积Kalman滤波；鲁棒滤波；信道丢包；网络攻击

 With the rapid development of information technology, the cyber-physical systems (CPS), which integrate the computing, communication and control capabilities, have received abroad attention and research. The state estimation system plays a key role to insure the stability and efficiency of CPS. However, due to the complexity and instability of the transmission network, the state estimation system in CPS faces the great challenges, such as limited bandwidth, packet dropout, cyber-attack, etc. To tackle these challenges, this thesis systematically addressed the event-triggered nonlinear estimation problem under the complex communication channel conditions. To reduce the amount of data transmission while ensuring the esstimation accuracy, an event-triggered cubature Kalman filter (ETCKF) is proposed. ETCKF uses the innovation based event-triggered sampling strategy in the sensor node to reduce the data transmission. Based on the nonlinear event-triggered strategy developed, the cubature Kalman flter (CKF), using the third-degree spherical-radial cubature rule, is adopted to further ensure the estimation accuracy. Further, the stochastic stability of ETCKF is analyzed. Using the stochastic Lyapunov stability lemma, ETCKF is proven to be stochastically stable if a sufficient condition, which is only composed of offline parameters, is satisfied. Moreover, the average communication rate of ETCKF is derived, which is only related to design parameters in innovation condition. To deal with the non-Gaussian or unknow noises, the stochastic event-triggered robust cubature Klaman filter (SETRCKF) is designed. Firstly, to make up for the deficiency of ETCKF, the stochastic event-triggered cubature Kalman filter (SETCKF) is proposed using the stochastic innovation based event-triggered sampling strategy, which can maintain the Gaussian property of the conditional distribution of the system state, and CKF. Based on SETCKF, SETRCKF is further designed by using the moving-window estimation method and adaptive method to estimate the measurement noise covariance matrices and the process noise covariance matrices and using the Huber fuction to make SETCKF more robust. Moreover, the stochastic stabilities of the two proposed filters are analyzed and the sufficient condition regarding the stochastic stability of the filtering error is derived. To tackle the packet dropout when using the stochastic innovation based event-triggered sampling strategy, the stochastic event-triggered cubature suboptimal filter (SETCF) is proposed. Firstly, by modeling the packet dropout as a Bernoulli process and inspired by the linear suboptimal filter, the cubature suboptimal filter (CF) is designed for periodic sampling system. Based on CF and the stochastic innovation based event-triggered sampling strategy, SETCF is further proposed. Moreover, the stochastic stabilities of the two proposed filters are analyzed by using the Lyapunov stability lemma. Considering that CPS is vulnerable to cyber attack and has limited bandwidth, the event-triggered cubature Kalman filters under two typical attack types, i.e., the data tampering attack and the deviation control command forgery attack, are established respectively. Aiming at the data tampering attack problem, the anomaly data detector is designed by using the projection statistics method. After the attack is detected, the weight matrix is accordingly constructed by using the detection result to correct the measurement value for the sake of ensuring the filtering accuracy, which complets the filter design. For the deviation control command forgery attack problem, the problem is firstly transformed into the problem that the system is with unknown input. Furthermore, the Bayesian inference method is used to derive the event-triggered cubature Kalman filtering algorithm. 关键词: Cyber-physical system; event-triggered sampling strategy; cubature Kalman filter; non-Gaussian noises; packet dropout; cyber attack.