This paper studies the multi-intelligent reflecting surface (IRS)-assisted cooperative sensing, in which multiple active IRSs are deployed in a distributed manner to facilitate multi-view target sensing at the non-line-of-sight (NLoS) area of the base station (BS). Different from prior works employing passive IRSs, we leverage active IRSs with the capability of amplifying the reflected signals to overcome the severe multi-hop-reflection path loss in NLoS sensing. In particular, we consider two sensing setups without and with dedicated sensors equipped at active IRSs. In the first case without dedicated sensors at IRSs, we investigate the cooperative sensing at the BS, where the target's direction-of-arrival (DoA) with respect to each IRS is estimated based on the echo signals received at the BS. In the other case with dedicated sensors at IRSs, we consider that each IRS is able to receive echo signals and estimate the target's DoA with respect to itself. For both sensing setups, we first derive the closed-form Cram\'{e}r-Rao bound (CRB) for estimating target DoA. Then, the (maximum) CRB is minimized by jointly optimizing the transmit beamforming at the BS and the reflective beamforming at the multiple IRSs, subject to the constraints on the maximum transmit power at the BS, as well as the maximum amplification power and the maximum power amplification gain constraints at individual active IRSs. To tackle the resulting highly non-convex (max-)CRB minimization problems, we propose two efficient algorithms to obtain high-quality solutions for the two cases with sensing at the BS and at the IRSs, respectively, based on alternating optimization, successive convex approximation, and semi-definite relaxation.

翻译：本文研究多智能反射面辅助的协同感知系统，其中多个主动式智能反射面以分布式部署方式，协助基站在其非视距区域实现多视角目标感知。与以往采用无源智能反射面的研究不同，本文利用具备信号放大能力的主动式智能反射面来克服非视距感知中严重的多跳反射路径损耗。具体而言，我们考虑两种感知配置：主动式智能反射面未配备专用传感器的情况，以及配备专用传感器的情况。在第一种无专用传感器的配置中，我们研究基站的协同感知机制，即基于基站接收的回波信号估计目标相对于每个智能反射面的到达方向。在第二种配备专用传感器的配置中，每个智能反射面能够独立接收回波信号并估计目标相对于自身的到达方向。针对两种感知配置，我们首先推导了目标到达方向估计的闭式克拉美-罗界。随后，在基站最大发射功率约束、各主动式智能反射面最大放大功率约束及最大功率放大增益约束条件下，通过联合优化基站的发射波束成形与多个智能反射面的反射波束成形，最小化（最大）克拉美-罗界。为求解由此产生的高度非凸的（最大）CRB最小化问题，我们针对基站端感知与智能反射面端感知两种场景，分别提出了基于交替优化、逐次凸逼近和半定松弛的高效算法来获得高质量解。