Top-down development of controllers for highly automated driving using solution spaces

Abstract

The complexity of highly automated driving functions and the large number of testing scenarios require the application of virtual design and testing methods. A common method to develop robust controllers is based on Γ-, Β and/or Θ- stability which need highly simplified vehicle models. Most of such models are linear and neglect relevant vehicle dynamics phenomena. Thus, it cannot be guaranteed that the closed-loop system including these simplified models covers all significant effects. This paper adopts a design method that computes solution spaces for controller parameters of arbitrary black box systems and models with no restrictions regarding linearity. This way the modelling represents the real system and disturbances more accurately. Parameter space approach and solution space approach are applied to a simple linear physical model of a steering system. Then, a more detailed model demonstrates the capabilities of the design method based on solution spaces. Requirements on the closed-loop system ex-pressed as boundaries of the stability regions are derived to ensure that the vehicle stays within a defined safety corridor next to the target trajectory. Additionally, a solution space for feasible controller parameters is specified. These requirements take nonlinear effects into account, thus increasing the validity of virtual testing. The effectiveness of this approach is demonstrated by designing the lateral controller of a highly automated vehicle.