Local Predictive Optimization of Globally Planned Motions for Truck-Trailer Systems

Logistics tasks often require the precise maneuvering of truck-trailer systems to a target position. For example, the rear of the trailer must be steered to a docking position for unloading. Global path planning algorithms usually generate suboptimal solutions. In an additional planning step, these solutions can be used as initialization for a local optimizer. In this article, a model predictive controller is presented, which performs a local optimization for suboptimal reference trajectories. Allowing to deviate temporarily from the previously generated solution allows to eliminate redundant direction changes, to increase obstacle clearance, and in the end, to generate more efficient solutions. To provide numerical robustness, it is proposed to constrain the terminal state to the reference path of the global planner solution. Simulation results for a one-trailer system highlight the benefits of the optimized solution in a particularly tight planning scenario. The applicability of the generated solutions is demonstrated in an experimental model scale setup. Furthermore, the method allows to switch between trajectory optimization (TO) and trajectory tracking (TT) mode to adapt to different driving situations, which is demonstrated in multiple simulation examples for a two-trailer system.