Item

Abstract

Motion cueing algorithms (MCAs) based on Model Predictive Control (MPC) are becoming increasingly popular. The MPC approach consists of solving an optimization problem to find a feasible simulator motion that minimizes the difference between the sensed motions in the real vehicle and in the simulator for some time interval. The length of this time interval, which is called the prediction horizon, is an important parameter that needs to be selected. Longer prediction horizons generally lead to better motion cueing but require more computational power because of the larger optimization problem. Consequently the selection of an appropriate prediction horizon for MPC-based MCAs is a compromise between motion cueing fidelity and computational load. In this work the effect of the prediction horizon on motion cueing fidelity was studied by computing the simulation cost, i.e., the average error between desired and reproduced sensory stimulation (specific forces and rotational velocities), for a range of typical car and helicopter maneuvers, while varying the prediction horizon. We propose a simple parametric model that describes the effect of prediction horizon on the simulation cost. The proposed model provides an accurate description of the data (coefficient of determination R2>0.99) for horizons longer than 1s for 11 out of 13 tested maneuvers. One of the model’s parameters can be interpreted as the minimal prediction horizon needed to achieve reasonable quality of simulation. The simulation cost appears to decrease roughly quadratically with the prediction horizon.