Fundamental Limits for ISAC-Asymptotics in Massive MIMO Sensing Systems

Asymptotic analysis is a common tool in statistics aiming at investigating the properties of an inference methodology as the number of observations grows to infinity. Even if the asymptotic regime cannot be achieved in real-world scenarios, its practical usefulness has been proved in an uncountable number of engineering applications. In the contest of ISAC, one of the brightest example is the Massive Multiple-Input-Multiple-Output (MMIMO) communications framework. The breakthrough brought by the MMIMO systems was in showing that, as the number of antenna elements grows to infinity, linear combining and precoding algorithms can mitigate the interference even in the presence of a partial knowledge of the communication channel. Inspired by this fundamental result, in this chapter we show that the massive (asymptotic) paradigm can bring essential benefits also in radar systems. In particular, we considered a co-located MIMO radar having a massive number of virtual spatial antenna channels. We focus on the target detection problem by showing that the massive regime allows for the derivation of a cognitive, robust, reinforcement learning (RL)-based, Wald-type test that guarantees certain performance regardless of the unknown statistical characterization of the disturbance. As concluding remarks, some explorative idea on a massive integrated communication/sensing system will be provided.