A novel computationally efficient algorithm in terms of the time-varying symbolic dynamic method is proposed to estimate the unknown initial conditions of coupled map lattices （CMLs）. The presented method combines symbolic dynamics with time-varying control parameters to develop a time-varying scheme for estimating the initial condition of multi-dimensional spatiotemporal chaotic signals. The performances of the presented time-varying estimator in both noiseless and noisy environments are analysed and compared with the common time-invariant estimator. Simulations are carried out and the obtained results show that the proposed method provides an efficient estimation of the initial condition of each lattice in the coupled system. The algorithm cannot yield an asymptotically unbiased estimation due to the effect of the coupling term, but the estimation with the time-varying algorithm is closer to the Cramer-Rao lower bound （CRLB） than that with the time-invariant estimation method, especially at high signal-to-noise ratios （SNRs）.
A novel approach to the inverse problem of diffusively coupled map lattices is systematically investigated by utilizing the symbolic vector dynamics. The relationship between the performance of initial condition estimation and the structural feature of dynamical system is proved theoretically. It is found that any point in a spatiotemporal coupled system is not necessary to converge to its initial value with respect to sufficient backward iteration, which is directly relevant to the coupling strength and local mapping function. When the convergence is met, the error bound in estimating the initial condition is proposed in a noiseless environment, which is determined by the dimension of attractors and metric entropy of the system. Simulation results further confirm the theoretic analysis, and prove that the presented method provides the important theory and experimental results for better analysing and characterizing the spatiotemporal complex behaviours in an actual system.
耦合映象格子(Coup led m ap lattice,CM L)模型是非线性科学研究中的一个重要模型。从CM L系统中恢复系统的初始条件对信号处理等问题的研究非常重要。本文在符号动力学方法的基础上,针对全局耦合映象模型,提出一种基于时变映象系数恢复信号初值的新方法。实验结果表明,本文方法能够有效地恢复初值的统计特性,使整个格点信号均值等于给定信号均值。此外,采用时变映象系数恢复的信号值与原信号之间偏差、均方误差(M ean square error,M SE)都较传统方法小,而且与原信号之间有较强的相关性,能够更好地刻画实际信号的物理过程,并对系统初始条件作出稳键估计。
