Goal-oriented Push-Pull Access: from DT alignment to ISAC

Federico Chiariotti [email protected] Fabio Saggese [email protected] Goal-oriented Push-Pull Access: from DT alignment to ISAC Future of Communications Workshop 2025 Two complementary approaches 2 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 AI-driven wireless sensor networks Let us consider an industrial IoT scenario with multiple nodes sensing multi-modal data How do we synchronize the Digital Twin? 3 [1] Pandey, S.R., Saggese, F., Chiariotti, F., et al. “Medium Access for Push-Pull Data Transmission in 6G Wireless Systems.” Submitted to IEEE Communication Magazine, online, 2025 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Pull-based goal-oriented communication 4 Pull-based communication allows the BS to get data from nodes that it knows that the DT has a high uncertainty over ➔No need for contention (all transmissions are scheduled centrally) ➔On average, information has a high value (reduced uncertainty) ➔Unexpected events may be missed [2] Gündüz, D., Chiariotti, F., et al., “Timely and Massive Communication in 6G: Pragmatics, Learning, and Inference.” IEEE BITS - The Information Theory Magazine, online, 2023 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Push-based goal-oriented communication 5 Pull-based communication allows the BS to get data from nodes that it knows that the DT has a high uncertainty over ➔Locally observed anomalies can be promptly reported ➔Nodes need to coordinate, and collisions are inevitable under high traffic loads [3] Chiariotti, F., Munari, A., et al., “Goal-Oriented Medium Access with Distributed Belief Processing.” IEEE/ACM Transactions on Networking, online, 2025 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Comparing the two approaches Push pros: ✓Event-driven approach: Enables immediate reporting of anomalies or time-sensitive data ✓Local view: Devices transmit based on actual observations ✓Decentralized: No need for constant coordination Push cons: ✗Local view: Devices cannot optimize for system-wide goals, as they only have local information ✗Uncoordinated access: Can lead to inefficient use of channel resources and a high collision risk Pull pros: ✓Coordinated access: Reduced collisions ✓Overall/global view: Central coordination improves system-wide resource allocation ✓System-driven relevance: The receiver can prioritize data based on global goals Pull cons: ✗Coordinated access: possible high overhead due to signaling ✗Overall/global view: decisions rely on predictions rather than measurements ✗Epistemic errors: data is only sent when requested, which may delay critical updates if anomalies occur 6 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Push-pull integration We can imagine a combined system which exploits synergies between the two approaches 7 When and how do we collect the data for pull-based processing? How do we provision resources for push-based access? [4] Chiariotti, F., Saggese, F.., et al., “A Combined Push-Pull Access Framework for Digital Twin Alignment and Anomaly Reporting.” Submitted to IEEE INFOCOM 2026, online, 2025 Integrated push-pull access 8 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Frame structure 9 F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Results coexistence Combination of all the benchmark schemes, considering the best values of Fixed P/Q compared with the dynamic subframe allocation strategy Perspective toward ISAC F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 What happens in a not static environment? Mobility scenarios require knowledge of communicating sensor locations to ensure accurate mapping of sensor data to the positions from which it is collected Can we use (random access) push or (scheduled) pull requests to obtain the sensor localization together with the sensory data? F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Fundamental limits for a single push packet Well… maybe! ■ A single snapshot (push packet) under MIMO-OFDM is decently good for range and Doppler Push for immediate ranging; Pull for tracking the movement of the sensors. F. Chiariotti, F. Saggese - Goal-oriented push-pull access: from DT alignment to ISAC - Future of Communication IV - 2025 Towards ISAC push/pull: challenges ■ How to create a meaningful object function merging the sensor data relevance and position uncertainty is an open question ■ Bistatic MIMO-OFDM models struggle with a non-singular formulation of the FIM if considering CTO and CFO. How to solve this is still a big issue (to me at least) ■ Considering both MAC and PHY layer model to have meaning results is hard to implement and evaluate the performance using a common metric is hard. Thank you for your attention!