TL;DR
This paper tackles the challenge of improving radar's ability to pinpoint targets in non-line-of-sight (NLOS) scenarios by jointly optimizing radar waveforms and the phase shifts of multiple intelligent reflecting surfaces (IRS). They introduce a novel algorithm, UBER, that efficiently designs these parameters to significantly reduce the uncertainty in estimating a target's direction-of-arrival (DoA).
What Problem Does This Paper Address?
Traditional radar systems struggle in environments where direct line-of-sight (LoS) to a target is blocked. Intelligent Reflecting Surfaces (IRS) offer a promising solution by acting as programmable mirrors, redirecting signals to illuminate targets indirectly. While previous work has focused on optimizing IRS for single-IRS radar systems, this paper addresses the more complex scenario of multi-IRS-aided MIMO radar. The core challenge is that to maximize the performance (specifically, the accuracy of target direction-of-arrival estimation), we need to simultaneously design both the radar's transmitted waveforms and the phase shifts of all the IRS elements. This joint optimization is computationally difficult due to the inherent constraints of unimodular waveforms (simpler hardware) and the NP-hard nature of the optimization problems.
Methods
The authors formulate the problem of minimizing the Cramér-Rao Lower Bound (CRLB) for target DoA estimation. They derive the Fisher Information matrix for this multi-IRS MIMO radar setup. To handle the complex joint optimization of unimodular waveforms and IRS phase shifts, they transform the problem into two Unimodular Quadratic Programs (UQPs). Since UQPs are NP-hard, they propose a novel algorithm called UBER (Unimodular Waveform and Beamforming design for multi-IRS radar). UBER leverages power method-like iterations (PMLI), a computationally efficient approach for approximating UQP solutions. The algorithm iteratively optimizes the waveform and the IRS phase shifts, effectively splitting the quartic optimization problem for phase shifts into two quadratic subproblems.
Datasets
Not discussed in the paper. The evaluation is based on simulated data.
Key Insights and Ideas
- Synergy of Multi-IRS: Deploying multiple IRS platforms, rather than just one, significantly enhances the radar's ability to estimate target parameters, especially in NLOS conditions.
- Joint Design is Crucial: Optimizing waveforms and IRS phase shifts together is key to unlocking the full potential of IRS-aided radar, leading to better performance than optimizing them independently.
- UBER Algorithm Efficiency: The proposed UBER algorithm, based on power method-like iterations, provides a computationally feasible way to solve the complex joint optimization problem, making it practical for real-world applications.
- Unimodular Waveforms for Hardware: The focus on unimodular waveforms is motivated by hardware simplicity and power efficiency, making the proposed solution more practical.
Key Results
Numerical experiments demonstrate that the UBER algorithm effectively improves the CRLB of DoA estimation. The paper shows that the multi-IRS-aided radar system, designed using UBER, outperforms single-IRS systems. Figure 1 in the paper visually confirms that increasing the number of IRS platforms leads to a lower CRLB (i.e., better estimation accuracy), and that the UBER algorithm achieves a monotonically decreasing CRLB with iterations.
Limitations and Future Work
The authors acknowledge that future work should consider enhancements such as accounting for inter-IRS interference and the quantization of IRS phases. They also mention that the analysis can be extended to 3D scenarios. The current work focuses on a specific target model (Swerling 0), and exploring other target models or more complex scenarios would be valuable.
Summary updated