In integrated sensing and communication (ISAC) systems, random signaling is used to convey useful information as well as sense the environment. Such randomness poses challenges in various components in sensing signal processing. In this paper, we investigate quantizer design for sensing in ISAC systems. Unlike quantizers for channel estimation in massive multiple-input-multiple-out (MIMO) communication systems, sensing in ISAC systems needs to deal with random nonorthogonal transmitted signals rather than a fixed orthogonal pilot. Considering sensing performance and hardware implementation, we focus on task-based hardware-limited quantization with spatial analog combining. We propose two strategies of quantizer optimization, i.e., data-dependent (DD) and data-independent (DI). The former achieves optimized sensing performance with high implementation overhead. To reduce hardware complexity, the latter optimizes the quantizer with respect to the random signal from a stochastic perspective. We derive the optimal quantizers for both strategies and formulate an algorithm based on sample average approximation (SAA) to solve the optimization in the DI strategy. Numerical results show that the optimized quantizers outperform digital-only quantizers in terms of sensing performance. Additionally, the DI strategy, despite its lower computational complexity compared to the DD strategy, achieves near-optimal sensing performance.

翻译：在集成感知与通信（ISAC）系统中，随机信号既用于传递有用信息，也用于感知环境。这种随机性给感知信号处理中的各个组件带来了挑战。本文研究了ISAC系统中用于感知的量化器设计。与大规模多输入多输出（MIMO）通信系统中用于信道估计的量化器不同，ISAC系统中的感知需要处理随机的非正交发射信号，而非固定的正交导频。综合考虑感知性能与硬件实现，我们聚焦于基于任务且受限于硬件的量化器设计，并采用空间模拟合并方法。我们提出两种量化器优化策略，即数据依赖型（DD）与数据无关型（DI）。前者通过较高的实现开销获得优化的感知性能。为降低硬件复杂度，后者从随机视角出发，针对随机信号进行量化器优化。我们推导了两种策略下的最优量化器，并基于样本平均近似（SAA）设计算法以求解DI策略中的优化问题。数值结果表明，优化后的量化器在感知性能上优于纯数字量化器。此外，尽管DI策略的计算复杂度低于DD策略，其仍能实现接近最优的感知性能。