Reconfigurable Intelligent Surfaces (RISs) have emerged as cost-effective technologies for improving wireless signal transmission. Conventional RIS designs, however, face challenges such as bulkiness, high production costs, limited scalability, and complex installation due to their reliance on wired connections. In this work, we introduce MetaMosaic, a novel RIS platform that repurposes 920 MHz RFID tags into battery-free unit cells, enabling an affordable, scalable, and flexible one-bit phase-modulated RIS. The system is engineered for compatibility with 2.4 GHz Wi-Fi communications while being controlled at 920 MHz. Our design incorporates two central innovations: the transformation of commercial RFID tags into functional unit cells and the development of a tailored neural radiance field to guide efficient reconfiguration. To further enhance global search capability and support multi-hotspot alignment, we extend the system with a genetic algorithm (GA)-based optimization strategy. Compared with the vanilla MetaMosaic, the GA-based MetaMosaic achieves an additional 3.1 dB signal strength improvement and enables simultaneous enhancement for up to five target points. Extensive testing across ten diverse environments demonstrates that MetaMosaic consistently boosts signal strength, with a mean gain of 19 dB over non-RIS setups. This outperforms current leading RIS systems by a 3-fold improvement.

Autonomous vehicles rely on Full Self-Driving (FSD) models for perception and trajectory planning. However, their wireless communication systems face significant challenges in dynamic environments, particularly evident in beamforming dependent vehicle-to-satellite links. Motivated by the powerful sensing capabilities of autonomous vehicles, we introduce Full Smart-Communication (FSC), a unified framework that reuses intermediate outputs from FSD pipelines—such as planned trajectories, occupancy maps, and 3D posture estimates—to model the spatio-temporal radio environment. To avoid accessing raw sensor data directly, FSC leverages processed semantic and kinematic information to anticipate channel conditions and proactively compute beamforming parameters ahead of time. Tested in realistic driving datasets, FSC improves signal strength by up to 15 dB. FSC bridges the gap between autonomous perception and robust communication, enabling proactive, low-latency, and energy-efficient connectivity.