Multi Stage Model Predictive Trajectory Set Approach for Collision Avoidance

Abstract

The presented approach combines the planning of trajectories and the vehicle control during emergency maneuvers. For this purpose an approach is utilized, which predicts the future behavior of the actuators and the vehicle with a nonlinear model. The input space is roughly discretized and a trajectory set is calculated explicitly. The choice of optimal inputs is performed by a direct comparison of the possible trajectories, in contrast to model predictive control. The discretization is carried out adaptively depending on the current reference input. Issues arising from the limited degree of freedom are solved by an additional transition time within the prediction horizon. Model inaccuracies are taken into account during the objective function evaluation, by utilizing a soft constraint function, which increase the distance to objects and street boundaries.