Stabilizing Model Predictive Control with Optimized Terminal Sample Weight for Motion Cueing Algorithm

Abstract

Driving simulators are cost-effective tools widely used in training, safety assessment, human factor research, virtual prototyping, etc. However, these devices suffer from workspace limitation which prevents a simulator driver from impression of a realistic driving. A Motion Cueing Algorithm (MCA) is used to produce the best motion sensation in a limited environment. Recently, Model Predictive Control (MPC) has been broadly used for MCA applications. One of the main challenges for MPC based MCA is to produce acceptable results in a real-time. A method to reduce computational burden is to decrease the prediction horizon. However, a very short prediction horizon will increase output fluctuation and unstable behavior of the output. In this paper, a terminal sample weighting method has been applied to stabilize an MPC-based MCA with a short prediction horizon and reduce the output error and undesired fluctuations. Then, a multi-objective genetic algorithm is applied to optimize the terminal sample weight in order to obtain the best realism of sensation. Computer simulations have verified the effectiveness of the proposed method in terms of human sensation improvement within the displacement restriction.