基于PSO-CNN-GRU模型的无人机短期航迹预测OACSTPCD
Short-term Track Prediction of UAV Based on PSO-CNN-GRU Model
针对无人机(Unmanned Aerial Vehicle,UAV)航迹预测问题,为了提升航迹预测的收敛速度和精度,论文提出了一种基于粒子群算法优化,卷积神经网络与门控循环单元网络相结合的PSO-CNN-GRU无人机航迹预测模型.为了解决神经网络人工调参难以获得最优解的问题,通过PSO算法进行自动调参,对GRU网络的隐藏层规模、学习率、批训练大小等参数进行优化,避免形成局部最优解;针对历史关键信息与重要特征的提取问题,通过CNN网络提取变量间的局部依赖关系,实现隐藏特征的挖掘.实验结果表明,与原始GRU模型相比,PSO-CNN-GRU模型的MAE、MSE的值分别降低了65.13%、73.25%,有着较好的准确性与鲁棒性.
Aiming at UAV track prediction,in order to improve the convergence speed and accuracy of UAV track prediction,a PSO-CNN-GRU UAV track prediction model based on particle swarm optimization and convolutional neural network and gate re-current unit network is proposed in this paper.In order to solve the problem that it is difficult to optimize the super parameters in the traditional cyclic neural network,the parameters of GRU network such as hidden layer scale,learning rate and training batch size are optimized automatically by PSO algorithm to avoid forming a local optimal solution.For the extraction of historical key informa-tion and important features,the local dependence relationship between variables is extracted by CNN network to realize the mining of hidden features.The experimental results show that,compared with the original GRU model,MAE and MSE values of PSO-CNN-GRU model are reduced by 65.13%and 73.25%respectively,which has good accuracy and robustness.
张成佳
陆军特种作战学院 桂林 541002
计算机与自动化
粒子群(PSO)算法卷积神经网络(CNN)门控循环单元(GRU)航迹预测无人机
particle swarm optimization(PSO)algorithmconvolutional neural network(CNN)gate recurrent unit(GRU)track predictionUAV
《舰船电子工程》 2024 (005)
45-49 / 5
评论