Joint Range-Angle Estimation for FDA–MIMO Radar: A Probabilistic Tensor Factorization Approach

This article addresses the unambiguous range-angle estimation problem in FDA-MIMO radar systems by formulating received signals as tensor-structured data and developing a tailored probabilistic tensor factorization algorithm. At the modelling stage, we firstly introduce how the compensated signal model of FDA-MIMO is linked to a complex-valued tensor decomposition problem from optimization perspectives. To employ a probabilistic framework, the optimization problem is then interpreted using probabilistic language. Finally, a tensor-based approach is proposed that jointly estimates the range and angle of targets utilizing variational Bayesian inference (VBI) techniques. Critically, our core innovation extends beyond a direct application of an existing Bayesian CP framework. We develop specific modifications to the standard Variational Bayesian Inference (VBI) procedure to handle the unique challenges of the FDA-MIMO signal model. These include the explicit formulation for complex-valued factor matrices with inherent Vandermonde structure and the derivation of efficient update rules that directly operate on this structure. Compared to the prevailing joint angle-range estimation techniques, the proposed method furnishes better estimation precision and more stable performance. Moreover, the algorithm remains effective even in situations where the number of targets is not known as a priori. At the analysis stage, a comprehensive analysis of its convergence properties, automatic tensor rank learning mechanisms, uniqueness condition, and algorithm complexity evaluation is provided. The simulation outcomes validate our theoretical dissection and high light the superior estimation precision of the proposed algorithm in contradistinction to extant schemes.