Stratospheric aerostats serve as cost-effective platforms for long-duration wireless communication at an altitude of 20 km.This study presents a receding horizon controller based on wind and balloon models to optimize the station-keeping performance.First,a neural network based on UNet is utilized for forecasting short-term wind fields.Then,an online receding horizon controller based on simultaneous optimistic optimization(SOO)is proposed for trajectory planning,which is particularly suitable for discrete action space.The SOO is modified and a reward function is designed to comply with restricted flight zone constraints and balance power consumption and station-keeping performance.Simulations across various positions and times demonstrate that the proposed method outperforms 90% and 40% cases separately compared to traditional greedy and A* algorithms,while make full use of the information of the wind.Besides,the reward function is effective to balance the station-keeping performance and the power consumption.The algorithm is potential for other optimal control problems with discrete action space.
维普
万方数据