Study On Robust Estimation Of Vehicle Motion States Utilizing Set-membership Filtering Theory

Reliable vehicle motion states are critical for the precise control performed by vehicle active safety systems.Acquiring vehicle motion states through soft-sensing method,which combines filter/observer theory with the information obtained by ordinary on-board sensors,can be taken as an alternative for the expensive and professional sensors due to its low cost and easy migration.However,most soft-sensing methods are based on the assumption that the system noise is the Gaussian white noise,which cannot describe the noise characteristics accurately for real systems.In addition,the accurate calculation of tire forces is difficult to be guaranteed for the strategy based on vehicle dynamics model.The factors mentioned above will seriously affect the robustness of the vehicle state estimation in complex real vehicle driving environment.In order to improve the robustness for the estimation of the longitudinal speed,lateral speed and sideslip angle,this thesis works from two aspects: the strategy construction of vehicle motion state estimation and the scheme design of tire force calculation.The main contents of this thesis are as follows:Firstly,the set-membership filtering theory is introduced into the field of vehicle motion state estimation.And it is mended via singular value decomposition technology to improve the numerical stability.For the set-membership filtering theory,the system noise is assumed as unknown but bounded,which is easy to be satisfied in real systems.Therefore,it can solve the defect that the traditional strategies can only obtain optimal estimation results when system noise is known as the Gaussian white noise.The improved algorithm is an important basis for the robust estimation of vehicle motion states.Secondly,the improved algorithm is applied to estimate tire forces.The influence of tire force on vehicle motion state estimation is confirmed by numerical simulation.Then a decoupling tire force estimation scheme is proposed.This scheme can be applied to complex driving conditions because it avoids the requirement of tire parameters and road conditions claimed by the traditional tire model-based methods.Simulation results show that the tire force calculation scheme is not sensitive to the change in road conditions,and can reach accurate and robust results.Finally,a robust estimation strategy for vehicle motion states is proposed,which combine the improved extended set-membership filter with a 3-degree-of-freedom vehicle model and the proposed tire force calculation scheme.Simulation results indicate that the strategy is robust to varying system noise.It has obvious advantages in the estimation performance of accuracy and real-time ability,and can provide guaranteed bounds for vehicle motion states.The semi-physical experiment results based on real vehicle data further prove the high-precision of the strategy in practical application.