A Driving Risk Prediction Method Based on Time-varying State Transition Probability Markov Chain

Abstract

The present invention provides a driving risk prediction method based on a time-varying state transition probability Markov chain. 1. Offline driving risk prediction model training: based on accident and nearby accident samples, real-time driving is divided by clustering time window characteristic parameters Risk status and treat it as a Markov chain listable state, establish a multi-logistic model of driving risk status transfer under different driving risk status; 2. Real-time prediction of online driving risk model: In the Internet of Vehicles environment, real-time collection of prediction models The required variable parameters are used to calculate the initial state probability distribution vector and the n-step transition probability of the Markov chain at any time in the future through the risk state cluster center position and Markov properties to obtain the future vehicle risk state prediction results. The present invention realizes the estimation of n-step time-varying state transition probability of the Markov chain through a recursive algorithm, which can reflect the characteristics that the driving risk state transition probability changes with the changes of traffic system characteristics, and can meet the real-time requirements of early warning.

Description

technical field [0001] The invention relates to the technical field of traffic safety evaluation and active safety of intelligent traffic systems, in particular to a driving risk prediction method based on a time-varying state transition probability Markov chain. Background technique [0002] On the basis of the perception of the running state of the vehicle and surrounding vehicles, research and prediction of the future driving risk state of the vehicle is helpful to realize the accurate and timely collision warning or intervention mechanism of the assisted driving system. At present, driving risk early warning mainly calculates the selected early warning variables in real time and compares them with preset risk thresholds of different levels. The widely used early warning variables mainly include collision time, inter-vehicle time and distance. In fact, it is difficult to describe the entire process from the formation of driving risk to the occurrence of dangerous conflict...

AI Technical Summary

Problems solved by technology

In fact, the entire process from the formation of driving risks to the occurrence of dangerous conflicts is difficult to describe with a single early warning parameter, and more complex models and algorithms are needed for research

However, the current early warning model algorithms at home and abroad usually only consider the operating characteristics of vehicles (such as inter-vehicle distance, speed and acceleration characteristics, etc.), while ignoring the impact of dynamic driver behavior, road and environmental changes on the driving risk state, and cannot fully reflect the driving characteristics. The internal evolution law between states is not conducive to the accuracy and prediction accuracy of the driving risk model

Images