Research Of The SUV Rollover Warning Based On Hidden Markov Model

In recent years, with the rapid development of automobile industry and road traffic, vehicle rollover accident has become an important safety issue more attention by people. In a relatively short period of time vehicle rollover is likely to happen, when the vehicle with a fast moving and emergency steering.so the vehicle rollover warning is particularly important.This article mainly have a research of the SUV vehicle's untripped rollover, using Hidden Markov Model(HMM) for rollover warning, and can real-time monitoring and have a prediction of vehicle motion state, in addition can issue the warning in advance, so as to improve the safety of the vehicle.First using the Carsim software simulation in the four working conditions which is prone to rollover,such as step steering,slope steering, doublelanechange steering and fishhook steering and can get the HMM's observed sequence: roll angle and lateral acceleration.For acquisition of data preprocessing, according to the motion state of vehicles: linear motion, normal steering, steering of emergency and rollover state,completing the data classification and segmentation in the four kinds of state, and using the K- means algorithm to determine the threshold of the motion state, as the premise of model training and recognition in this paper.Second build the double HMM motion state model, the ground floor of the model is multidimensional vehicle motion parameters, the top of the model corresponds to the motion state of Multidimensional Gaussian Hidden Markov Model(MGHMM).And training model, based on the Baum-Welch algorithm to identify data from the complex working conditions, to identify the current state of vehicle. At the same time using the Markov prediction method, to predict what will happen in the future 3s of the vehicle, in case of a rollover then trigger the early warning device,else over the prediction cycle.Finally the Artificial Neural Network(ANN) combined with the HMM, with the HMM currently used to identify the state of the vehicle motion parameters value as input of ANN model, and to train the ANN model of BP neural network algorithm to predict the value of the three motion parameters : roll Angle, lateral acceleration,steering wheel of the next period of time. HMM is achieved to predict the vehicle's motion state of the next period of time, and ANN has realized the data value parameters' prediction of vehicle next time, a combination of the two model, can make drivers more intuitive and specifically to determine the degree of danger of vehicle rollover when it happens, at the same time the parameters of the ANN forecasting value can be as the input reference for the upper rollover control system, so as to make the whole system more integrated.