This work introduces STCIB, the first Set Transformer-based beamforming framework for cell-free integrated sensing and communication (CF-ISAC). Instead of relying on computationally heavy iterative optimization or locally constrained learning architectures, the method models the CF-ISAC network as an unordered set and uses global self-attention to capture interactions among distributed access points, users, and sensing targets.
The resulting architecture is permutation-invariant, operates in an unsupervised manner, and produces beamforming solutions through a single forward pass. In this view, the method transforms estimated channels into set-structured features, propagates information through attention blocks, and outputs a power-constrained beamforming matrix suitable for sensing-centric, communication-centric, and joint ISAC regimes.
Paper: Set Transformer-Based Beamforming Design for Cell-Free Integrated Sensing and Communication.