Answers are grounded in this paper and cite their sources.
Conventional automotive radars can detect obstacles in direct line of sight, while partially or fully occluded scene elements are difficult to detect. In this paper, we investigate the deployment of intelligent reflecting surfaces (IRS) to help automotive radars enhance non-line-of-sight (NLOS) visibility, thus reducing blind spots and improving safety. Our novel approach relies on passive beamforming using IRS to detect multiple targets in NLOS regions by maximizing the power reflected from a designated radar area toward the target regions. We formulate this IRS configuration problem as a pattern synthesis task by matching the reflective beam pattern with a desired pattern, and we solve it using the gradient descent algorithm. Subsequently, for a given IRS configuration, we tackle the target location estimation problem by posing it as a sparse recovery task and solve it using a sparse Bayesian learning algorithm. Numerical simulations demonstrate that our approach achieves improved target localization performance in comparison to a random IRS approach.
No takes yet. Share an insight, caveat, or question.
Chakraborty et al. (2025) studied this question.