Model Predictive Control for Agricultural Machines with Implements

This paper presents a method to stabilize a specific point on a vehicle along a given reference path or trajectory. To this end, based on the classical kinematic single-track model, the rear axle midpoint is transformed to any point that is rigidly coupled to the vehicle, e.g., an agricultural implement. An invariant tracking error is formed which is required for trajectory tracking and path-following using a nonlinear model predictive control. This provides the possibility to react to kinematic restrictions and actuator dynamics as well as to delay times in the actuator system. The underlying optimization problem is solved using a gradient-based augmented Lagrangian approach in order to achieve real-time feasibility. The algorithms are validated in simulations. A hardware-in-the-loop simulation is performed on an embedded electronic control unit to prove real-time capability.