Dynamic resource allocation optimization algorithm for massive MIMO communication based on Bayesian optimization

Dynamic resource allocation in massive MIMO systems faces severe challenges such as high dimensionality, non-convexity, multiple constraints, and rapid time-varying channel states. The two-stage Bayesian optimization algorithm can effectively address these challenges by constructing a Kriging surrogate model and introducing a Portfolio acquisition function selection strategy, enabling intelligent and efficient joint optimal allocation of power and bandwidth resources in high-dimensional black-box spaces. Based on this, a dynamic resource allocation optimization algorithm for massive MIMO communication under Bayesian optimization is studied. With the goal of maximizing system energy efficiency, a joint optimal allocation model including power, bandwidth, and beam resources is constructed under constraints such as total power, subcarrier power, and bandwidth. The Kriging model is introduced as the probabilistic surrogate model for Bayesian optimization to approximate the complex relationship between the objective function and constraints in the allocation model. Combined with the Portfolio strategy to adaptively select the acquisition function, the selected acquisition function adaptively approaches the global optimal allocation scheme through two stages of search and allocation. Experimental results show that with 256 antennas configured in the MIMO system, the maximum system throughput after allocation by the algorithm reaches 14.5 Mbps; with 128 antennas, the system energy efficiency can reach a peak of 445 Mbits/Joule, which can intelligently balance spectral efficiency and circuit power consumption. In multi-user and dynamic scenarios, the algorithm also exhibits excellent interference management capabilities and robustness, providing an efficient and reliable optimal allocation scheme for precise resource allocation of massive MIMO systems in complex dynamic environments.