交通运输系统工程与信息 ›› 2021, Vol. 21 ›› Issue (1): 90-94.

• 智能交通系统与信息技术 • 上一篇    下一篇

考虑航艏向与数据变化差异的船舶轨迹预测

高天航*1,徐力1,靳廉洁1, 2,葛彪1   

  1. 1. 交通运输部规划研究院,水运所,北京 100028;2. 大连海事大学,交通运输工程学院,辽宁 大连 116000
  • 收稿日期:2020-11-11 修回日期:2020-12-06 出版日期:2021-02-25 发布日期:2021-02-25
  • 作者简介:高天航(1991- ),男,辽宁海城人,工程师,博士。

Vessel Trajectory Prediction Considering Difference Between Heading and Data Changes

GAO Tian-hang*1, XU Li1, JIN Lian-jie1, 2, GE Biao1   

  1. 1. Division of Waterway Planning, Transport Planning and Research Institute Ministry of Transport, Beijing 100028, China; 2. College of Transport Engineering, Dalian Maritime University, Dalian 116000, Liaoning, China
  • Received:2020-11-11 Revised:2020-12-06 Online:2021-02-25 Published:2021-02-25

摘要:

船舶自动识别系统(Automatic Identify System,AIS)数据可以实时体现船舶当前时刻的具体动态,采用传统BP(Back Propagation)神经网络模型的船舶轨迹分析预测方法,在计算中直接将航艏向数据纳入模型,没有考虑船舶航艏向在零度附近变动时带来的实际方向变动幅度与数据变化幅度存在较大偏差问题。为解决该问题,在BP神经网络基础上,引入双三角函数变换,同时将正弦值与余弦值纳入模型,将两者相结合,从两维度体现航艏向情况;在拟合预测后进行反三角函数变换和平均处理,构建一种基于改进神经网络算法的船舶AIS轨迹预测模型。选取实例数据进行模型验证,实例结果表明,该模型预测结果比不考虑差异方法的误差均方差更小,大幅降低误差幅度,可更精确地预测船舶轨迹。

关键词: 水路运输, 船舶轨迹预测, BP神经网络, AIS数据, 三角函数

Abstract:

Automatic identification system (AIS) data can reflect the specific dynamic of the ship at the current moment in real time, and the existing BP (Back Propagation) neural network based methods for ship trajectory analysis and prediction only take the heading data into the model directly. The methods do not consider the large deviation between the actual direction change range and the data change range when the ship heading changes near zero. In order to solve this problem, a ship AIS trajectory prediction model based on the improved neural network algorithm is constructed in this paper. The model introduces the double trigonometric function transformation on the basis of BP neural network. The sine value and cosine value are included in the model to consider the two-dimension direction of the heading. The inverse trigonometric function transformation and average processing are carried out to postprocess the predicted data. By selecting the case data to verify the model, the case results show that the prediction error of the model is smaller than the method without considering the difference, which greatly reduces the error range and can be more accurate for ship trajectory prediction.

Key words: waterway transportation, vessel trajectory prediction, BP neutral network, AIS date, trigonometric function

中图分类号: