A Dynamic Weight-Based Load Balancing Initial Access Algorithm for LEO Mega-Constellations

Abstract

Although Low Earth Orbit (LEO) megaconstellations offer diverse initial access choices via multisatellite coverage, uneven spatiotemporal traffic frequently causes MAC-layer collisions and physical-layer disconnections. To address this, we propose an intelligent cross-layer initial access algorithm using dynamic comprehensive weights. By jointly normalizing visible elevation angles and real-time satellite loads, our algorithm employs a threshold-driven adaptive mechanism that smoothly shifts optimization priorities between maximum link quality and strict congestion avoidance. Evaluated via a high-fidelity discrete-event simulator integrating real TLE ephemeris and a Markov-based user traffic model, the strategy effectively disperses bursty requests. Results demonstrate that under extreme concurrency, the proposed algorithm overcomes both single-point collisions of greedy strategies and link outages of pure load balancing, achieving a superior 92% access success rate while maintaining robust physical-layer signal-to-noise ratios (SNR) and minimizing global load imbalance.