Active 2-D DOAs and Reflection Coefficient Estimation Using Special Partitioning and Nystrom via Massive MIMO Radar

Modern radar systems for applications such as autonomous navigation, surveillance, and smart sensing increasingly rely on massive multiple-input multiple-output (MIMO) architectures to achieve high-resolution estimation of target direction and reflectivity. However, the joint estimation of two-dimensional (2D) direction-of-arrival (DOA) angles (azimuth and elevation) in such systems is computationally intensive, particularly when conventional methods rely on optimizations, high-dimensional spectral searches, or full matrix eigendecomposition. The techniques recently presented in the literature offer improvements but remain limited in scalability and real-time feasibility. This paper presents a low-complexity framework for active 2D DOAs and target reflection coefficient estimation tailored for co-located massive MIMO radar systems. The approach, called the special partitioning and Nystrom-based approach (SPNA), restructures received data through strategic vectorization to decouple azimuth and elevation estimation and leverages array geometry to partition the problem into smaller, tractable subspaces. The Nystrom method is applied to approximate large eigendecompositions using only partial array data, enabling significant computational savings without sacrificing accuracy. SPNA avoids multi-dimensional searches and cumbersome pairing operations through a novel pairing-free 1D search scheme. Simulation results demonstrate that SPNA accurately resolves closely spaced targets and achieves estimation performance competitive with the state-of-the-art, while significantly reducing the computational burden, particularly when leveraging parallel processing. The framework has potential for real-time radar applications and is scalable to future massive MIMO deployments.