A Vehicle Rollover Warning Approach Based on Neural Network and Support Vector Machine.

Abstract: This paper studies the vehicle rollover warning based on the vehicle running state parameters which are easy for collecting. By taking advantages of neural network and support vector machine, this paper analyzes the probabilities of the vehicle rollover in the scenario of single lane change and establishes a warning model which could predict the rollover 0.4seconds in advance for a particular type of vehicle. According to the validation experiments on Ve- DYNA which is a vehicle dynamic simulation software, the results validate the warning model''s good performance with the maximum error as 0.04 second and no missing cases which indicates its validity for rollover warning. [Copyright &y& Elsevier]