Collaborative Sensing Research Of Multi-information For Vehicle System And Its Simulation Verification Based On Distributed Structure

At present,the development of autonomous vehicle is vigorously boosting the research of advanced driving assistance systems and active safety systems.For the stable operation of the vehicle,it is of great significance to perceive various kinds of the required information accurately.Most of the existing methods for information perception can only obtain one kind of information,and/or they are often work in a specific and simple scene,such as level road and unchanged vehicle mass.However,there are often coupling relationships between different types of vehicle information.And the way of perceiving one type of information alone is often encountering issues of robustness and applicability.In addition,due to terrain restrictions or design requirements,there are usually slopes on the road,and the vehicle mass also changes with load.Under the effect of all these factors,there is a huge risk that the existing methods will fail.Aiming at these problems,this thesis tries to develop a high-performance acquisition method for multi-information including the vehicle motion states,vehicle mass and road slope.The main content can be divided into the following aspects.Firstly,a multi-information collaborative sensing scheme based on the distributed structure is proposed for obtaining different types of information.The inner information circulation inside this scheme is implemented by the coupling relationships among multi-information,and the perception of different types of information is realized by three modules: vehicle motion state estimation,vehicle mass determination and road slope observation.Secondly,considering the impact of road slope on vehicle dynamics,a mathematical model is established to describe this situation accurately.On this basis,a vehicle motion state estimation module is designed by employing the variational Bayesian-based adaptive cubature Kalman filter algorithm.In the time-variant noise environment,this algorithm can jointly estimate the system states and measurement noise,and obtain the local optimal solution by minimizing the KullbackLeibler divergence.Then,the vehicle mass determination module is designed based on the forgetting factor recursive least squares algorithm and vehicle longitudinal dynamics model.To reduce the time delay in the road slope observation results,a slope observation module is constructed by using a novel slope prediction model and a cubic observer.The comparison results show that the vehicle mass accuracy of the proposed module is higher and the time delay of the slope results can also be alleviated effectively.Finally,the work of this thesis is verified and compared via experiments.The results confirm that in the time-variant measurement noise environment,the perception result of the vehicle motion state estimation module is more accurate and has better robustness.Moreover,when the slope changes and the vehicle mass is unknown,the proposed collaborative sensing strategy has obvious advantages of low time delay,high precision and good stability.