LAUSR

In this article a motion cueing algorithm (MCA) based on model predictive control (MPC) for a hexapod-based dynamic driving simulator is derived. The design objective of the MCA is to reproduce the real world accelerations and angular velocities for a test person in the simulator while respecting its actuator limitations. This results in an underlying nonlinear, state-constrained optimal control problem. In order to exploit the full predictive potential of the described algorithm, a method to predict the future driver behavior and thus the future desired values for the MPC is derived by modeling the driver as an optimal controller. The OCP weights that mainly influence the predicted driving actions are learned from demonstration using an inverse optimal control (IOC) approach. Furthermore, the direct incorporation of human perception models into the motion planning process is considered. An online driver-in-the-loop experiment with the Daimler driving simulator shows the high potential of the derived MPC scheme compared to the commonly used filter-based approach as well as the efficiency of the underlying optimization method.