TL;DR
This paper introduces a novel "Unitary ESPRIT" algorithm for simultaneously estimating the direction of departure (DOD) and direction of arrival (DOA) of multiple targets in a Multi-Overlapped-Transmit-Subarray (MOTS) MIMO radar system. The proposed method leverages rotational invariance properties and a centro-Hermitian matrix transformation to achieve a closed-form solution, offering improved accuracy and reduced computational complexity compared to existing techniques.
What Problem Does This Paper Address?
MIMO radar offers significant advantages over traditional radar, including enhanced resolution and a larger effective aperture. However, a key challenge is the loss of transmit directional gain when transmitting orthogonal waveforms omni-directionally, which degrades the signal-to-noise ratio (SNR) per virtual antenna and impacts angle estimation accuracy. Subarray MIMO radar, particularly the Overlapped Transmit Subarray (OTS) configuration, aims to mitigate this by coherently processing signals within subarrays, thus improving SNR.
While existing angle estimation techniques like ESPRIT have been applied to MIMO radar, they often suffer from high computational costs due to exhaustive peak searches or limitations in handling specific configurations. Specifically for bistatic MIMO radar, joint estimation and automatic pairing of DOD and DOA for multiple targets remain challenging. This paper addresses the need for an accurate, computationally efficient, and robust method for joint DOD and DOA estimation in the MOTS MIMO radar configuration, which combines the benefits of subarray processing with advanced estimation techniques.
Methods
The core of this paper is the development of a "Unitary ESPRIT" scheme tailored for the MOTS MIMO radar. The approach involves several key steps:
- Partitioning OTSs: The K Overlapped Transmit Subarrays (OTSs) are divided into two groups that exhibit a rotational invariance relationship. This is a crucial step for applying ESPRIT-like techniques.
- Synthetic Observation Data Matrix: A synthetic observation data matrix (Z) is constructed. This matrix is designed to have its data samples doubled by exploiting the properties of centro-Hermitian matrices.
- Real-Valued Transformation: The centro-Hermitian matrix Z is transformed into a real-valued matrix (Ω) using a unitary transformation. This step is significant because real-valued computations are generally less complex than complex-valued ones.
- Eigenvalue Decomposition (EVD): The real-valued matrix Ω is subjected to an eigenvalue decomposition (or Singular Value Decomposition - SVD) to extract the dominant left singular vectors.
- Rotational Invariance Exploitation: The rotational invariance relationship is then applied to the real-valued manifold to derive a closed-form solution for the angles. This involves manipulating specific block selection matrices and their properties.
- DOA and DOD Estimation: The algorithm derives a closed-form solution for both DODs and DOAs, and importantly, these estimates are automatically paired. The DODs are obtained through an arcsin transformation of eigenvalues from an intermediate matrix, while DOAs are derived from the difference between adjacent elements of the estimated steering vectors.
The "unitary" aspect refers to the use of unitary transformations to convert complex-valued data to real-valued data, which simplifies computations and can improve estimation accuracy.
Datasets
Not discussed in the paper. The evaluation is performed using simulated data.
Key Insights and Ideas
- Leveraging MOTS Configuration: The paper effectively exploits the unique structure of MOTS MIMO radar, where overlapped subarrays provide coherent gain, leading to improved SNR per virtual element. This is a key differentiator from traditional MIMO radar.
- Centro-Hermitian Transformation for Real-Valued Processing: The clever use of a centro-Hermitian matrix and its subsequent transformation to a real-valued matrix is a significant algorithmic contribution. This reduces computational complexity and potentially improves numerical stability.
- Automatic Pairing of DOD and DOA: A critical practical advantage is the automatic pairing of DOD and DOA estimates, eliminating the need for computationally expensive matching procedures often required in joint estimation algorithms.
- Trade-off in OTS Number: The simulations highlight an optimal number of OTSs (K=3 or 4) that balances effective aperture and SNR gain per virtual antenna, demonstrating that simply increasing the number of subarrays isn't always best.
Key Results
- Improved Accuracy: The proposed Unitary ESPRIT scheme for MOTS MIMO radar demonstrates increased estimation accuracy compared to traditional MIMO radar employing ESPRIT and Unitary ESPRIT, especially at lower SNR regions.
- Reduced Computational Complexity: The method's computational complexity is shown to be O(μM²K²P), which is approximately equal to unitary ESPRIT in [17] and lower than the ESPRIT method in [15]. This is attributed to the real-valued processing and efficient matrix manipulations.
- Lower SNR Threshold: The proposed method exhibits a lower SNR threshold for successful detection, meaning it can achieve reliable angle estimation at lower signal strengths.
- Performance vs. Number of Pulses: The accuracy improves with an increasing number of pulses, but the improvement plateaus, suggesting that the MOTS MIMO's inherent SNR gain is a primary driver of performance.
- Optimal Number of OTSs: Simulations indicate that an optimal number of OTSs (K=3 or 4) provides the best performance, balancing effective aperture and SNR gain.
Limitations and Future Work
- Assumptions: The method relies on assumptions such as targets being in the same range bin and the use of uniform linear arrays (ULAs). The performance with non-ULA configurations or targets in different range bins is not explored.
- Ideal Conditions: The derivation and simulations are based on ideal conditions, such as known waveform orthogonality and the absence of mutual coupling between antennas. Real-world scenarios might introduce deviations.
- Future Work: The paper implicitly suggests that further exploration of the optimal trade-off between effective aperture and SNR gain for different MOTS configurations could be a direction for future research. Investigating the algorithm's performance with more complex target scenarios (e.g., closely spaced targets, extended targets) would also be valuable.
Summary updated