Vehicle track destination prediction method considering space-time semantics and driving state

Abstract

The invention discloses a vehicle trajectory destination prediction method considering space-time semantics and a driving state, and the method comprises the steps: firstly extracting the position semantics of each departure region and a middle trajectory point based on an interest point data set and a TF-IDF algorithm; secondly, obtaining departure time information from a timestamp of a starting track point, and processing the departure time information by adopting heuristic cyclic fuzzy coding to obtain monthly, weekly and daily multilevel departure time semantics; then, extracting a driving state sequence, and splicing the driving state sequence with a track point sequence, position semantics and departure time semantics to form an input feature sequence; and then, constructing a deep learning prediction model, learning a long-term dependency relationship of an input feature sequence by using a double-layer LSTM, designing a space-time attention mechanism to capture key travel space-time features, and mapping the learned space-time features into predicted destination coordinates through a multi-layer full-connection residual network. The method provided by the invention can greatly improve the prediction precision of individual travel driving destinations.

Description

technical field [0001] The invention relates to the fields of individual travel mode modeling and vehicle trajectory prediction, in particular to a vehicle trajectory destination prediction method that takes into account space-time semantics and driving status. Background technique [0002] Movement trajectory destination prediction is an important branch in the research field of human travel patterns. It aims to predict the most likely destination of the trip based on a movement trajectory that has not yet completed the travel process and the corresponding travel context information. Vehicle trajectory destination prediction has been widely studied on public transportation trajectory data. With the development and popularization of mobile positioning technology, the research on travel patterns at the individual level of non-public vehicles will become a hot spot, and for personalized service recommendation, traffic navigation, vehicle Location-based services such as insuran...

AI Technical Summary

Problems solved by technology

[0007] In view of this, the present invention provides a vehicle trajectory destination prediction method that takes into account space-time semantics and driving state, to solve or at least partially solve the technical problem of low precision existing in the methods in the prior art

Images