首页|期刊导航|飞控与探测|基于LSTM和迁移学习的无人飞行器目标意图预测方法

基于LSTM和迁移学习的无人飞行器目标意图预测方法OA

Method for Aircraft Target Intention Prediction Based on LSTM and Transfer Learning

中文摘要英文摘要

无人飞行器由于其小型化、无人化、多功能化的特点,被广泛应用于多种任务场景中.针对传统深度学习方法在小样本数据量下,无法快速、准确地预测无人飞行器意图等问题,提出基于长短时记忆(Long Short-Term Memory,LSTM)网络和迁移学习的无人飞行器意图预测方法,使用根据专家经验建立的虚拟数据集,将虚拟数据集做SG滤波处理后送入LSTM神经网络中,得到预训练模型,最后使用实际飞行试验小样本数据,迁移该预训练模型,得到最终LSTM和迁移学习模型.实验结果表明,LSTM和迁移学习模型在小样本数据下,预测准确率得到提升.

Unmanned Aerial Vehicles are widely used in various task scenarios due to their minia-turization,unmanned nature,and multifunctionality.To address the issue where traditional deep learning methods cannot quickly and accurately predict UAV intentions under small sample data,a method using Long Short-Term Memory(LSTM)networks and transfer learning is proposed for UAV intention prediction.A virtual dataset,established based on expert experience,is prepro-cessed using SG filtering before being input into the LSTM neural network to obtain a pre-trained model.Finally,the pre-trained model is transferred to actual small sample flight test data,resul-ting in the final LSTM and transfer learning model.Experimental results show that the LSTM and transfer learning model improve prediction accuracy under small sample data.

谈用杰;邵昱轩;陈力恒

哈尔滨工程大学智能科学与工程学院·哈尔滨·150001哈尔滨工程大学智能科学与工程学院·哈尔滨·150001哈尔滨工程大学智能科学与工程学院·哈尔滨·150001

信息技术与安全科学

无人飞行器LSTM网络迁移学习SG滤波小样本数据集

unmanned aerial vehicleLSTM networktransfer learningSG filteringsmall sample data-set

《飞控与探测》 2026 (4)

62-68,7

国家自然科学基金(62273114)黑龙江省自然科学基金(YQ2024F015)中央高校基本科研业务费专项资金资助(3072025YC404)黑龙江省春雁支持计划(CYQN24071)

10.20249/j.cnki.2096-5974.2026.04.005

评论