A Scheme of Gradient-Based Iterative Identification With Separable Synchronous Variable-Innovation for Exp-ARX Systems With Arbitrary Data Loss

This article addresses the parameter identification of the Exp-ARX system with arbitrary data loss. To effectively tackle the challenges associated with data loss to enhance data utilization and identification accuracy, a novel iterative method for constructing output predictions is proposed, which operates in multiple steps starting from the most recently previous output value. To reduce the computational complexity caused by numerous characteristic parameters and further improve the accuracy of parameter estimation, an iterative algorithm based on separable synchronization is proposed by using the hierarchical principle. The key problem of this algorithm is to transform the original system into two subsystems and estimate their parameters separately. For each subsystem, a gradient-based iterative algorithm with variable innovation is introduced to adjust the length of innovation data adaptively. The performance of the proposed algorithm has been verified through simulation and applications, demonstrating significant improvements in handling data loss and parameter estimation accuracy.