Fast Data Aggregation based on Over-the-air Computation
Yun Chen, Yongling Wu, Na Yang · 2025
Because of the maturity and development of wireless communication technology and artificial intelligence (AI), application scenarios requiring multi-participant cooperation continue to emerge. utilizing the superposition characteristics of wireless communication channel, Over-the-air Computation (AirComp) has become an excellent fast and secure wireless data aggregation method. As far as we know, the peak power is usually as the constraint for the power optimization design of the system in most of the research work on AirComp. In this study, we recommend a dynamic power allocation strategy for edge devices to optimize the aggregation error of AirComp. To further overcome the fading characteristics of wireless signals, an Intelligent Reflector Surfaces (IRS) assisted AirComp is proposed to construct a controllable wireless communication environment. Unfortunately, the resulting optimization problem is a non-convex quadratic constrained quadratic programming (QCQP). We adopt the alternating Difference of Two-Convex Functions (D.C.) to deal with the QCQP problem. The data results show that compared with the current solution, the dynamic power distribution between users can reduce the system power, and the system performance can be further promoted by using IRS assisted communication.