基于强化迭代学习的分布式无人机编队控制研究

Research on Distributed Unmanned Aerial Vehicle Queue Control Based on Reinforcement Iterative Learning

摘要:

在无人机行驶过程中,基站主机所定义的队列信息会对与飞行器相关的转向行为造成一定的影响,故为保证无人机飞行器的稳定行驶状态,基于强化迭代学习对分布式无人机队列控制算法展开研究。计算强化学习函数的具体数值,通过迭代处理的方式,实现对迭代值概率系数的分布表示,完成强化迭代学习模型的设计。以此为基础,定义无人机队列拓扑结构,并求解信息迁移指标的具体数值,实现基于强化迭代学习的无人机队列信息迁移。在无人机队列控制器的配合下,建立分布式队列信息集合,并联合其中的队列数据样本,求解UAV控制参数。再根据行进队列建模条件,完善控制算法执行流程,完成基于强化迭代学习的分布式无人机队列控制方法的设计。实验结果表明,在强化迭代学习模型的影响下,无人机转向角始终保持在0°-90°的数值范围之内,表示飞行器按照基站主机所定义的队列信息行驶,能够始终保持较为稳定的运动状态,符合实际应用需求。

Abstract:

In the process of UAV driving, the queue information defined by the base station host will have a certain impact on the steering behavior related to the UAV. Therefore, in order to ensure the stable running state of the UAV, the distributed UAV queue control algorithm is studied based on reinforcement iterative learning. Calculate the specific numerical value of the reinforcement learning function, and through iterative processing, achieve the distribution representation of the probability coefficient of the iterative value, and complete the design of the reinforcement iterative learning model. Based on this, define the topology structure of the drone queue, and solve the specific numerical value of the information transfer index to achieve the drone queue information transfer based on reinforcement iterative learning. With the cooperation of the UAV queue controller, the distributed queue information set is established, and the queue data samples in it are combined to solve the UAV control parameters. Then, according to the modeling conditions of the travel queue, the control algorithm execution process is improved, and the design of the distributed UAV queue control method based on enhanced iterative learning is completed. The experimental results show that under the influence of the reinforcement iterative learning model, the steering angle of the unmanned aerial vehicle remains within the numerical range of 0 ° -90 °, indicating that the aircraft can always maintain a relatively stable motion state according to the queue information defined by the base station host, which meets the practical application requirements.