RiLoCo: An ISAC-oriented AI Solution to Build RIS-empowered Networks

TL;DR

This work addresses joint sensing and communication optimization in RIS-enabled networks for 6G by introducing RiLoCo, an ISAC-oriented framework. It combines a Fisher-information-based localization metric with a RIS-driven coverage metric into a single objective $M_S(\boldsymbol{d}) = \alpha L_S(\boldsymbol{d}) + (1-\alpha) C_S(\boldsymbol{d})$ and uses Q-learning to efficiently search the deployment space under a budget. The key contributions include a novel deployment formulation that accounts for environmental shadowing and device heterogeneity, a projection-based localization framework, a unified joint metric, and an iterative RL-based solver that yields near-optimal performance for both sensing and communication. The approach demonstrates substantial practical impact by enabling energy-aware, high-precision localization with robust coverage, guiding real-world RIS installations for future ISAC-enabled networks.

Abstract

The advance towards 6G networks comes with the promise of unprecedented performance in sensing and communication capabilities. The feat of achieving those, while satisfying the ever-growing demands placed on wireless networks, promises revolutionary advancements in sensing and communication technologies. As 6G aims to cater to the growing demands of wireless network users, the implementation of intelligent and efficient solutions becomes essential. In particular, reconfigurable intelligent surfaces (RISs), also known as Smart Surfaces, are envisioned as a transformative technology for future 6G networks. The performance of RISs when used to augment existing devices is nevertheless largely affected by their precise location. Suboptimal deployments are also costly to correct, negating their low-cost benefits. This paper investigates the topic of optimal RISs diffusion, taking into account the improvement they provide both for the sensing and communication capabilities of the infrastructure while working with other antennas and sensors. We develop a combined metric that takes into account the properties and location of the individual devices to compute the performance of the entire infrastructure. We then use it as a foundation to build a reinforcement learning architecture that solves the RIS deployment problem. Since our metric measures the surface where given localization thresholds are achieved and the communication coverage of the area of interest, the novel framework we provide is able to seamlessly balance sensing and communication, showing its performance gain against reference solutions, where it achieves simultaneously almost the reference performance for communication and the reference performance for localization.

Figures (8)

• Figure 1: -assisted network simultaneously considering both indoor and outdoor scenarios.
• Figure 2: Schematic view of the elements of the system model.
• Figure 3: Time of arrival (ToA) (left) and angle of arrival (AoA) (center) measurements diagrams in 2D, with their combination (right).
• Figure 4: Overview of RiLoCo building blocks.
• Figure 5: Example of a base scenario, with CSs, one BS and shadowing areas.

Theorems & Definitions (1)

• Theorem 1