A Method for Identifying Most Significant Vehicle Parameters for Controller Performance of Autonomous Driving Functions. 2019-01-0446
This paper is about the identification of most significant vehicle parameters influencing the behavior for a lateral control system for autonomous driving. Requirements for a control function for autonomous driving need to consider many uncertainties in the plant. While most uncertainties can be compensated by application, some vehicle properties can change significantly during usage. Performance measures, which are influenced by the changing vehicle properties can decide if a control system is admissible. The analyzed parameters are functional tire characteristics, mass and position of center of gravity. Since the parameters are dependent, but Sobol sensitivity analysis assumes decoupled inputs, random variation yields no reasonable results. Furthermore, if each parameter or set of parameters would be variated individually the dimensionality of inputs and with it the numbers of simulations would increase significantly. Therefore, the proposed methodology analyses the influence of the parameters on performance measures of the lateral control function for autonomous driving. First, a hierarchical nonlinear principal component analysis using artificial neural networks is used to decouple inputs and reduce them to a lower dimensional subspace. In a second step Sobol sensitivity analysis is used to identify the significant parameters by sampling in the decoupled subspace and full vehicle simulation. The most significant parameters are merged in a static and dynamic variable resulting in four worst-case vehicle states. Those vehicle states are fitted by an extended single-track model. Performance measures of the lateral control function as rising time, overshoot, number of oscillations can be analyzed for robustness with each vehicle model separately using cooperative design.
Jan-Dominik Korus, Markus Bullinger, Christoph Schuetz, Pilar Garcia Ramos, Sebastian Wagner, Steffen Müller
BMW AG, Technische Universität München, Technische Universität Berlin