The evolution of wireless communications is moving towards increasingly intelligent and efficient systems. One of the most promising developments is the combination of reconfigurable intelligent surfaces (RIS) with pinching antenna systems (PASS), an architecture that promises to significantly improve coverage, capacity and energy efficiency in multi-user environments. However, the joint optimization of antenna positions, RIS phase shifts and beamforming vectors presents a computational challenge of high complexity. In this context, graph neural networks (GNNs) emerge as an effective solution, capable of learning complex patterns in network topologies and delivering real-time approximations. This article analyzes from a technical and business perspective how GNN optimization is transforming these systems and how companies can leverage these innovations through the development of custom software and artificial intelligence solutions.
RIS-assisted PASS systems combine two disruptive technologies. On one hand, pinching antennas allow the physical displacement of radiating elements along a waveguide, adjusting their position to maximize the signal towards users. On the other hand, RIS modify the electromagnetic environment through tunable reflective elements, creating additional virtual channels. The integration of both introduces a non-convex optimization problem with multiple constraints: the movable region of the antennas, the total power budget and the adjustable phases of RIS elements. Traditionally, convex optimization methods, such as interior point algorithms or decomposition techniques, offer optimal solutions but with high computational costs, limiting their real-time applicability. This is where GNNs make a difference: by modeling the network as a graph —where nodes represent antennas, users and RIS elements, and edges represent electromagnetic interactions— the GNN learns in an unsupervised way to predict near-optimal configurations with minimal latency. Recent research, such as the reference paper arXiv:2511.20305v2, proposes a three-stage GNN that first learns antenna positions, then phase shifts and finally beamforming vectors, achieving a balance between speed and optimality.
From a business perspective, this technology opens opportunities in sectors such as telecommunications, Industry 4.0 and private network deployments. Implementing a GNN-based solution for RIS-PASS optimization requires robust software capable of integrating machine learning models with embedded systems and physical device control. Here, Q2BSTUDIO, as a software and technology development company, offers services ranging from the creation of custom artificial intelligence to cloud infrastructure deployment. For example, training complex GNN models requires scalable computing environments on platforms like AWS or Azure, where optimization algorithms can run with large volumes of simulated data. Additionally, cybersecurity plays a crucial role in sensitive wireless networks; a PASS-RIS system can be vulnerable to phase injection attacks or malicious interference, so integrating cybersecurity solutions is essential to ensure communication integrity.
Another relevant aspect is the monitoring and performance analysis of these systems. Companies need dashboards that visualize key metrics such as sum rate, energy efficiency or inference latency. Business Intelligence (BI) solutions based on Power BI allow integrating data from simulations or real deployments, generating dynamic reports that help engineers make informed decisions. Q2BSTUDIO develops custom dashboards that connect with cloud databases and offer interactive visualizations, adapting to the specific requirements of each project. Likewise, process automation through AI agents can optimize the reconfiguration of RIS in real time, responding to changes in user mobility or channel conditions.
The impact of key parameters, such as the number of RIS elements or transmission power, has been studied in depth. Numerical results show that GNNs not only achieve near-optimal theoretical performance, but also generalize well to unseen scenarios during training, thanks to the topological structure of the graph. This is especially valuable in dynamic environments where user locations vary constantly. For a company looking to implement this technology, the first step is to have a development team capable of designing the GNN model, training it with synthetic or real data and deploying it on an embedded system or in the cloud. Here, Q2BSTUDIO's capabilities in multi-platform application development come into play, both for the RIS controller side and for the user interface that allows operators to monitor the network.
In summary, the integration of RIS and PASS optimized by GNN represents a qualitative leap towards sixth-generation (6G) wireless communications. Companies that adopt these technologies early will be able to offer services with higher capacity, lower energy consumption and better user experience. The key lies in combining knowledge in communication theory with solid skills in software development, artificial intelligence and cloud computing. Q2BSTUDIO provides the necessary technical support to address these challenges, from conceptualization to final implementation, ensuring that innovation translates into real competitive advantages.
