Research On Intelligent Vehicle Lane Changing Behavior Judgment And Collision Risk Estimation

Intelligent vehicles, which can effectively improve the driving safety and reduce traffic accidents, are the future development trend of automotive technology. The anti collision technology of intelligent vehicles is an important technical means to improve the safety of automobile. This paper focuses on the problem of lane changing behavior determination and collision risk estimation of anti collision technology, the purpose is to determine whether the vehicle is in a lane changing state and the collision risk estimation method. The research of this paper has certain theoretical and practical significance to the driving safety of the intelligent vehicle.This paper is divided into four chapters, lane changing behavior determination and collision risk estimation method of intelligent vehicles are studied. A model which describes vehicle lane changing behavior as well as the training method of model parameters is given. The collision risk estimation method based on the driving track of the host vehicle and relative motion parameters between the front and host car are discussed. The feasibility of the method is analyzed through the experimental data.A Gaussian Mixture Hidden Markov Model is established in this paper based on the satisfaction of driving behavior and Markov process, and the deterministic and stochastic characteristics of lane changing behavior. The decision states of driver during a lane changing process is segmented and described by the model parameters while the lateral distance of the front vehicle and the center of the host vehicle measured by sensors are used to characterize the changes of driving decision states. An Expectation-Maximization training method is proposed to train the model parameters, and the validity of the method is verified by simulation.In order to test the effectiveness of the Gaussian Mixture- Hidden Markov Model and the training method of model parameters, a vehicle data acquisition system consists of a millimeter wave radar and camera sensors is constructed. The results of lane changing behavior judgment of front vehicles are given under the conditions of highway. The experimental results show that the proposed model and method can judge the lane changing behavior of adjacent vehicles effectively and timely.A prediction model of vehicle track is established in this paper in order to estimate the risk of collision between the front of the vehicle and host vehicle. The calculation method of collision probability between two cars based on Monte Carlo sampling method considering the uncertainty of measurement of relative motion parameters between the host car and front car is designed. The front of the vehicle is simulated by high precision vehicle dynamics software veDYNA and the risk assessment method was validated under the simulation conditions. The results indicate that the method proposed in this paper can effectively reflect the risk of collision between two cars.