Multi-fleet driving danger prediction system and method for reducing jitter

Abstract

The invention discloses a multi-fleet driving danger prediction system and method for reducing jitter. The prediction method comprises the steps: acquiring vehicle state feature information within a period of time through a data acquisition module; enabling the data analysis module to identify the driving behavior state through a time recurrent neural network LSTM and determine the final driving behavior state in combination with a time delay function; determining a behavior dynamic model of each vehicle according to the driving behavior recognition result, and calculating the actual acceleration of the vehicle through the vehicle following dynamic model; and determining the driving danger probability of all related vehicles in various driving behavior states based on a Monte Carlo simulation method and an importance sampling strategy. By combining the time recurrent neural network LSTM and the time delay function, various driving behaviors and driving strategies can be quickly, effectively and stably recognized and subjected to risk assessment, and the risk assessment accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of intelligent transportation, and relates to a multi-team driving risk prediction system and method for reducing jitter. Background technique [0002] In recent years, multi-vehicle cooperative driving system, as one of the main application scenarios of vehicle-road cooperative driving system, has received extensive attention from most research departments and enterprises. Relevant studies have proved that cooperative driving system can effectively improve traffic operation efficiency and Safe and smooth car driving. [0003] At present, the research and application direction of fleet cooperative control strategy is mainly focused on the relatively simple fixed single fleet control strategy, and the research purpose is to improve the driving safety and stability of a single fleet. Most of the research on cooperative driving focuses on the optimization of vehicle control stability and driving safety at the ...

AI Technical Summary

Problems solved by technology

Most of the research on cooperative driving focuses on the optimization of vehicle control stability and driving safety at the level of automatic control, but does not study the driving behavior modeling and driving state prediction in the case of mixed driving behaviors, so it is difficult to accurately evaluate the current situation. Dangers of Driving Behaviors and Strategies

Considering the complexity and variability of the actual driving environment and the different purposes of the drivers in the fleet, the conversion of driving behavior and the mixing of different driving behaviors will cause a large change in the driving state of the vehicle (even need to leave the current fleet), which will seriously affect the driving of the vehicle. Safe and easy to cause confusion in the surrounding driving environment

[0004] Prior art 1 (a multi-vehicle driving behavior analysis and danger warning method and system) discloses a driving behavior recognition model based on a neural network and a variable time window to determine the driving behavior category, but the time window is an estimated value, making it difficult to drive accurately Behavior category; prior art 1 uses the Gaussian model based on the error ratio and the Monte Carlo method to perform related simulations. It is only a theoretical introduction or conception of a sub-module introduced in the early warning system, and no actual discussion is made. The relevant calculation model has not been refined

Although importance sampling is an important strategy in the Monte Carlo method to reduce variance, the inventors found in actual research that the two factors involved in operation delay and fluctuation period are not suitable to be represented by a uniform distribution probability density function probability

Existing technology 1 uses the least square method to estimate the parameters of the high-order polynomial model, and the recognized driving behavior state has a large degree of jitter, which greatly reduces the prediction accuracy

Images