Research Of Rollover Warning System For Heavy Vehicles Based On Hidden Markov Model

The heavy vehicle has a higher gravity-center, large loads and large volume, whichmakes the dynamic roll stability of heavy vehicle poorly. In the emergency turning orchanging lanes case, it leads rollover accidents easily. In this paper, the research ofrollover early warning system for heavy vehicle based on reading a lot of literature onheavy vehicles rollover warning system and combining with The National Science Fundfor Young Scholars " A new algorithm Research of Rollover Warning System of HeavyVehicles Based on Hierarchical Hidden Markov Model ",it aims to establish a heavyvehicle rollover warning model, which can be real-time monitoring of the vehiclemotion state and early warning the vehicle in a rollover state which requires to takemeasures to avoid rollover.In this paper, the double-layer HMM model structure of the vehicles the movementgesture recognition and the vehicle motion state identification and rollover earlywarning for vehicle are established based on the hierarchical structure of HMM model.The underlying model corresponding to the vehicle motion gesture of MGHMM andhigh-level model corresponds to the complex conditions vehicle motion state MGHMMof the double-layer HMM model structure. The high-level and low-level combine acompleted layer model. The required training data is obtained though using theTrucksim simulation in the Step input, Ramp steering, Fishhook steering and Doublelane change four steering conditions. The noise data is eliminated smoothly by using theSavitay-Golay filtering method and the normal date in the short time period datecollection is removed. With the help of the off-line training method for vehicle motionMGHMM and vehicle motion state of the MDHMM is trained offline and get doubleHMM model parameters. The vehicle motion state boundaries value is set by usingk-means algorithm.In this paper, the potential risk of the vehicle rollover is predicted by using Markovprediction method. The vehicle emergency steering is treated as triggering the warningcondition. If the warning conditions are met and begin to predict the motion state of thevehicle within3seconds, if the prediction of the vehicle rollover dangerous state, anearly warning device is triggered and remind the driver to slow down or modifieddriving operation to prevent vehicle rollover occurred.The off-line identification of the vehicle motion posture and the vehicle motion state and offline validation validation of the vehicle rollover early warning are based onTrucksim, Simulink and Matlab toolbox of HMM and off-line training double HMMmodel parameters. It shows that HMM model early warning system is proposed in thispaper can be real-time monitoring of vehicle motion state, and early warning for vehiclerollover, and meet the real-time and accuracy requirements, and rollover warning alsoachieve good results by the identification of vehicle motion state and offline verificationresults of rollover warning in the complex conditions.