Joint Pilot and Data Power Optimization in Cell-free Massive MIMO Systems using Heterogeneous Graph Convolutional Network

Pilot contamination caused by the reuse of pilot sequences poses a significant challenge in cell-free massive MIMO systems. The problem can be mitigated by the optimization of pilot power to enhance channel estimation. In this paper, we propose a max-min achievable uplink rate optimization problem to jointly optimize powers allocated to pilots and data across all users. To tackle the intricate non-convex problem, we present a novel architecture based on a heterogeneous graph convolutional network. Simulation results demonstrate superior performance of our proposed algorithm compared to conventional approaches.