一种基于IPSO优化ELM的内河船舶轨迹预测方法

An IPSO-optimised ELM-based Method for Inland Vessel Trajectory Prediction

文中针对现有预测模型精度和计算效率不高的问题,提出一种引入深度学习框架(ELM)和改进粒子群算法(IPSO)的预测模型,进行内河船舶轨迹预测.采用长江武汉段实测AIS数据并从中提取一系列典型轨迹,使用ELM方法对典型轨迹进行实时预测,考虑到ELM网络的超参数难以人工取到最优,提出采用IPSO算法对网络超参数进行寻优处理,利用最优网络参数构建轨迹预测模型(IPSO-ELM),并与SVM方法、LSTM方法和ELM方法进行对比实验.结果表明:IPSO-ELM有效提高了轨迹预测精度.

Aiming at the low accuracy and computational efficiency of the existing prediction model,a prediction model with deep learning framework(ELM)and improved particle swarm optimization(IPSO)was proposed to predict the trajectory of inland river ships.A series of typical trajectories were extracted from the measured AIS data in Wuhan section of the Yangtze River,and the typical trajectories were predicted in real time by ELM method.Considering that it is difficult to get the optimal parameters of ELM network manually,an IPSO algorithm was proposed to optimize the network parameters.The trajectory prediction model(IPSO-ELM)was constructed by using the optimal network parameters,and compared with SVM method,LSTM method and ELM method.The results show that IPSO-ELM effectively improves the trajectory prediction accuracy.