SC²I Distinguished Lecture | Robust Over-the-Air-Aggregation: Integrated Communication and Computation for Future 6G Systems

Traditional paradigm has communication and computation separated. For example, in a sensor network with K devices, radio resource and multiple access schemes are used to coordinate the transmission so that the server / base station first detect the symbols transmitted by each devices and then compute the sum. Many existing multiple access schemes (such as collision avoidance schemes or contention resolution schemes) are designed based on this philosophy. However, O(K) radio resource will be needed to maintain a constant latency. In future applications such as remote state estimation or federated learning, the goal at the receiver is not to detect individual symbols from devices but rather, to compute the sum of the symbols. In this case, it is beneficial to integrate communication and computation. The author has proposed over-the-air aggregation as a new multiple access scheme based on analog transmission for this purpose. However, naïve analog transmissions will suffer from interference. Traditional error correction coding is incompatible with analog transmissions and they cannot be applied to enhance robustness against interference, In this talk, we propose a new method to enhance the robustness of detection for over-the-air-aggregation multiple access using sparse recovery methods. We show that the proposed solution has low complexity and superior performance compared with naïve analog transmission strategies.