基于计算图与CRLB的集群编队队形优化方法

Optimization of Swarm Formation Based on Computational Graph and CRLB

集群内定位精度是无人机集群应用于各类复杂任务的先决条件。面向集群编队队形定位优化需求,建立了基于相对距离和到达角的集群内协同融合定位模型,推导了领航-跟随编队下的融合定位克拉美罗下界理论极限值,建立了队形定位优化评价函数,作为集群编队队形优化理论依据。针对集群编队队形优化评价函数梯度求解难的问题,构建了队形定位优化评价函数的梯度计算图,通过反向传播,逐步得到最终梯度值,基于共轭梯度法,设计了最优队形搜索算法。通过数值仿真验证结果表明:与遗传算法和蜜獾算法相比,基于计算图与CRLB的集群编队队形优化方法可以得到更优的编队队形,集群理论定位精度也更高。

Intra-cluster localization accuracy is a prerequisite for UAV swarm to be applied to various complex missions.Facing the demand for optimization of formation positioning in cluster,a cooperative fusion positioning model based on relative distance and angle of arrival is established,the theoretical limit of the lower boundary of the fusion positioning Cramér-Rao lower bound under the leader-follower is derived,and the evaluation function of formation positioning optimization is established,which serves as the theoretical basis for the optimization of formation in cluster.Aiming at the problem that it is difficult to solve the gradient of the evaluation function of formation optimization for cluster,the gradient calculation graph of the formation positioning optimization evaluation function is constructed,and the gradient value is obtained step by step through back propagation,and the optimal formation search algorithm is designed based on the conjugate gradient method.Numerical simulation results show that compared with the genetic algorithm and the honey badger algorithm,the method proposed can obtain formation with higher cluster theoretical localization accuracy.