Two-Level Model-Based Control of Flexible Multibody Systems

We describe procedures for the simulation of maneuvers with multibody models. A maneuver is here mathematically defined as an optimal control problem. This definition feeds into a planning layer, that determines the vehicle trajectory by solving the optimization problem for a reduced vehicle model. The resulting trajectory in turn feeds into a control layer, that tracks the computed optimal path using a non-linear model predictive formulation and steers the multibody model accordingly. The reduced models used at the planning and tracking levels are augmented with a neural element that is adapted on-line to optimize the predictive capabilities of planner and tracker. The result is a system that not only is capable of piloting a vehicle model of arbitrary complexity along a given track, but it is also able to compute the track itself based on given criteria. The procedures are demonstrated on emergency maneuvers of rotary wing vehicles.