An automatic lane-changing intention labeling method based on high-noise vehicle trajectory data

Abstract

The present invention relates to a method for automatically marking lane-changing intentions based on high-noise vehicle trajectory data, comprising the following steps: S1: collecting and preprocessing the tracking trajectory data of surrounding vehicles; S2: searching for lane-changing points; S3: starting point of lane-changing OK; S4: valid lane change trajectory verification; S5: output the marked lane change frame and lane keeping frame. The present invention uses the time average change of the distance between the vehicle and the road center point as a marker for judging the start of lane changing, which not only considers the influence of noise, but also provides a robust definition for the nature of the ever-changing lane changing behavior.

Description

technical field [0001] The invention belongs to the technical field of unmanned vehicles, and in particular relates to a method for automatically marking lane-changing intentions based on high-noise vehicle trajectory data. Background technique [0002] With the development of unmanned driving technology, in order to safely and efficiently pass through complex traffic scenes and make optimal decisions, unmanned vehicles need to have the ability to predict the future intentions and trajectories of surrounding vehicles. The driving intentions of surrounding vehicles not only have an important impact on the decision-making planning of unmanned vehicles, but they also affect each other's driving intentions. Based on their own experience, human drivers can predict their driving intentions and future trajectories based on the current and past behaviors and interactions of surrounding vehicles, so as to make important driving decisions such as overtaking, slowing down or changing l...

AI Technical Summary

Problems solved by technology

Some of the existing learning-based algorithms use manual labeling of public datasets (such as NGSIM US101), which requires a lot of labor, and such public datasets are generally collected using road fixed sensors, which are different from those collected on unmanned vehicles. The data varies greatly in terms of accuracy, etc.; part of the data generated by simulation is used as a training set, which is not practical for practical application

[0005] A topic titled "Learning Vehicle Cooperative Lane-changing Behavior from Observed Trajectories in the NGSIM Dataset." (authors Su, Shuang & Mülling, Katharina & Dolan, John & Palanisamy, Praveen & Mudalige, Priyantha) published in 2018 In the public article, the automatic labeling method is used to label the vehicle intention of the NGSIM dataset, such as Image 6 As shown, the change of the angle between the vehicle orientation and the road is used as the sign of the start of the lane change, but this is unrealistic in practice. It has the following two problems: 1. In practice, the angle between the vehicle and the road is changed 2. And this method needs to fix a threshold for judging the change of the included angle, but the position and speed of each vehicle when changing lanes are different, so it is difficult to identify the change of the included angle under the tracking noise. road conditions do not apply

Images