Joint optimization of multi-beam configuration and resource allocation for low-earth orbit satellites

• Establishing a mixed-integer non-convex optimization model enables the simultaneous optimization of beam design, pointing angle, user association, and power allocation, thereby capturing the interrelationship between space and power. • A hierarchical iterative framework is proposed, decomposing the problem into user association, beam configuration, and power allocation to efficiently approach the global optimum. • An innovative beam pattern-pointing optimization algorithm is proposed, using exchange matching to dynamically optimize beam direction and coverage, enhancing system capacity. Traditional multi-beam satellite systems typically employ fixed beam pointing and static directional designs, lacking the ability to adapt spatial-domain resources. Consequently, service load is often distributed unevenly among beams or users, causing substantial mismatches between requested and allocated capacities. With the rapid evolution of Space-Air-Ground Integrated Networks (SAGIN) and the convergence of 5G/6G communications, flexible payloads and reconfigurable beam technologies have emerged as promising approaches to improve capacity utilization and service fairness. Motivated by these advances, this paper introduces a unified beam-configuration optimization framework that jointly integrates beam-pattern and beam-pointing design.We proposed a hierarchical iterative optimization method to address beam configuration, user association, and power allocation simultaneously. The overall coupled optimization problem was decomposed into three synergistic subproblems: beam configuration, user association, and power allocation. Specifically, the outer layer used an exchange-matching mechanism for dynamic beam-user mapping, while the inner layer employed a successive convex approximation (SCA) algorithm for efficient power allocation. Simulation results demonstrated that the proposed framework substantially improved capacity-demand matching, achieving 10–25% higher capacity than optimizing beam pointing or pattern alone, with robust performance, fast convergence, and superior energy efficiency across diverse user distributions.