From Pixels to Qubits: Quantum-Classical Fusion for Blockage Prediction in RIS-Aided Networks

Reconfigurable intelligent surfaces (RIS) offer programmable control of wireless propagation, but classifying user blockage remains challenging in dynamic environments. This paper formulates a quantum--classical fusion approach combining RGB images and RIS-assisted channel observations for ternary link-status classification: absent, blocked, or unblocked. Visual features are amplitude encoded, and scalar channel information can be represented through a qubit rotation. A variational circuit and classical readout map these inputs to class probabilities. We distinguish quantum encoding of classical observations from the stronger hypothesis of a physical quantum metasurface. The latter requires an explicit photonic implementation and cannot be established using classical wireless data alone. The framework motivates controlled comparisons with classical vision-only, channel-only, and multimodal models. We clarify state-preparation overhead, the dimensionality of the encoded states, and the limited interpretation of fidelity as a state-similarity measure. The illustrative plots retained in this review version are not verified experimental evidence. Claims of 99\% accuracy, quantum advantage, and superior runtime are therefore withheld pending traceable data, implementation details, and repeated-run evaluation.