Adaptive HAPS Offloading: A Strategy for Supporting RAN During High Traffic Load

High altitude platform stations (HAPS) have been proposed to support terrestrial mobile networks, offering a sustainable alternative to network densification. With their wide coverage areas and green energy consumption model, HAPS super macro base stations (SMBSs) are well suited to handle the massive and dynamic mobile data traffic demand. This research introduces an adaptive traffic offloading strategy that leverages the capabilities of HAPS to support radio access network (RAN), particularly during periods of high network demand. To enable HAPS to effectively assist the RAN, it is crucial to accurately predict which base stations (BSs) will experience high loads. Precise forecasting of these demands is hence essential to ensure timely and targeted offloading of traffic to the HAPS when and where it is most needed. The proposed approach predicts and manages loads by considering temporal and geographical factors. At the core of this approach is the Q-learning update rule, which is continuously used to refine offloading decisions and flexibly adapt to changing conditions. Our simulation results demonstrate that the proposed HAPS offloading approach is effective in maintaining balanced loads in the terrestrial RAN during peak periods, by dynamically adapting to the typical traffic characteristics of different areas and to their evolution over time.