Complex scene driving risk prediction method based on multiple time-space diagrams

Abstract

The invention provides a complex scene driving risk prediction method based on multiple space-time diagrams, and the method comprises the steps: constructing the multiple space-time diagrams which describe different space-time relationships among vehicles in a peripheral multi-vehicle complex scene, inputting the fused multiple space-time diagrams into a graph convolutional neural network, and extracting the feature vectors of the multiple space-time diagrams of the scene; taking a multi-time-space diagram feature vector sequence extracted at each moment in an observation period as a multi-step input feature of a long-short-term memory neural network, and training multi-vehicle time-space sequence samples in different risk states to obtain a driving risk prediction model; and acquiring motion information of the vehicle and multiple surrounding vehicles in real time, extracting a multi-space-time diagram feature vector sequence among all vehicles in an observation period in real time, inputting the multi-space-time diagram feature vector sequence into the driving risk prediction model, and finally obtaining a multi-vehicle collision risk state prediction result at a future moment. The method is suitable for multi-vehicle collision risk prediction in a peripheral multi-vehicle complex scene, and the accuracy and practicability of the prediction model are improved.

Description

technical field [0001] The invention relates to the technical field of traffic safety evaluation and intelligent traffic system active safety, in particular to a complex scene driving risk prediction method based on a multi-temporal-space graph. Background technique [0002] Collision risk estimation algorithm is one of the cores of smart car active safety technology. The performance of risk estimation algorithm will directly determine the timeliness and reliability of system early warning or active intervention, so it is also one of the key research contents of current automobile manufacturers and researchers. . The traditional collision risk estimation algorithm mainly quantifies the risk level of different driving conditions through the indicators that characterize the initial conflict state, that is, the risk estimation indicators are calculated based on the motion parameters of the two vehicles at the initial moment of the scene, and the calculated values ​​of the risk ...

AI Technical Summary

Problems solved by technology

However, these indicators are usually only applicable to the prediction of two-vehicle collision risk in specific longitudinal or horizontal scenarios, and generally do not consider the multi-vehicle collision risk problem in the complex scene of surrounding multiple vehicles that needs to be faced in the actual driving process, which is not conducive to the application of such methods. Application in real complex driving scenes

At the same time, since the collision accident is a small probability event, it is often difficult to obtain and divide the state of multi-vehicle collision risk samples, which makes it more difficult to build a multi-vehicle collision risk prediction model

Images