Deep learning-based method for building position prediction model by considering vehicle driving influence factors in internet-of-vehicles complex network

Abstract

Characteristics of criss-cross roads, non-uniform vehicle distribution and the like objectively exist in city roads, so that vehicle positions are easily changed to cause the problem of data transmission distortion of an internet-of-vehicles network layer, and the problem becomes a bottleneck hindering the development of internet-of-vehicles application services. An existing vehicle position prediction model is trained by generally utilizing historical track data of vehicles, so that consideration of complex vehicle states and real-time road condition information is lacking, and relationshipsbetween complex driving environments and vehicle driving behaviors and between the complex driving environments and vehicle position changes are mined insufficiently. For the problem, a deep learning-based method for building a position prediction model by considering vehicle driving influence factors in an internet-of-vehicles complex network comprehensively considers the vehicle driving influence factors such as vehicle body attributes, road information, driving environments and the like; in combination with a deep learning technology, the relationships between the vehicle driving influencefactors and the vehicle positions are mined; and the vehicle position prediction model is proposed, so that the purpose of improving vehicle position prediction accuracy is achieved, and assistance isprovided for improving the stability of route protocol design of the internet-of-vehicles network layer and effectively solving the data distortion problem.

Description

technical field [0001] The invention relates to the complex network field of the Internet of Vehicles. Background technique [0002] Existing vehicle position prediction mainly serves applications related to the safety of the Internet of Vehicles (such as vehicle anti-collision monitoring at intersections) or road traffic flow analysis. The prediction model mainly mines the historical trajectory and speed of the current vehicle, and uses the relationship obtained by mining to predict the position of the vehicle at the next moment. The researchers assume that the speed and acceleration of the vehicle will be constant in a short period of time in the future, and use the current speed and acceleration to calculate the position after the Δt time using the physical formula, and then calculate the position deviation. Lanes with more vehicles have larger deviations. The VMP probabilistic model proposed by the researchers divides the vehicle's driving path into several connected r...

AI Technical Summary

Problems solved by technology

[0003] The methods discussed above are too ideal for road modeling, ignoring many factors that may affect the position of vehicles, such as traffic signals and lane turning permission signs, etc. etc.) on driving behavior and vehicle position

Images