Model Predictive Reference Generation of Wheel-Individually Controlled Vehicles

Over-actuated vehicles featuring all-wheel inde-pendent drive and all-wheel independent steering with extended steering angles are able to achieve benefits for logistic and transportation systems due to the improved maneuverability. For such multi-wheeled systems, the paper at hand proposes a model predictive reference generation (mprg) of the steering angles and the rotational speeds of the vehicle's wheels. As an example system, a four-wheel drive and four-wheel steered electric vehicle is discussed. Based on the state of the art as well as previously published results [1], requirements of the algorithm are specified comprehensively. Hence, the proposed algorithm satisfies hard constraints due to limited steering angles and considers the actuator dynamics leading to smooth references for the underlying low-level controllers. Further-more, wear and tear of the tires are reduced and desired final states can be specified, such as turning the wheels back to their initial directions during parking procedures. From a control perspective, control signals given by motion control as well as control allocation are tracked without restrictions on the wheel's translational velocities. Therefore, the weightings of the algorithm's cost function are time-varying depending on the vehicle's velocity and an estimation of the wheel's current slip. Results are discussed with a slalom and a parking maneuver requiring steering angles that are larger as achieved by a conventional vehicle as well as the ISO double lane change challenging the velocity independence of the approach.