A study on state-of-the-art motion cueing algorithms applied to planar motion with pure lateral acceleration – comparison, auto-tuning and subjective evaluation on a KUKA robocoaster serial ride simulator

Driving simulators are widely used during the training of drivers/pilots, and for entertainment, as well as in the research of human behavior. Motion cueing algorithms (MCAs) are aimed at mapping the motion of moving vehicles into the limited workspace of driving simulators, while preserving the perceptual realism of the simulation. Several MCAs have been developed in the literature to improve the realism of the simulation. However, each MCA presents strengths and weaknesses when compared to the others. Most importantly, the tuning of the MCA parameters is an open issue because these parameters are not intuitive for normal simulator users. The current dissertation considers a systematic comparison of existing MCAs based on a simple maneuver with only lateral (left/right) accelerations, i.e. a trajectory along a planar S-shaped track with constant velocity. All the MCAs were implemented and compared numerically using a novel measure – the well-tuned index. In a later stage, the trajectories of the MCAs were implemented in the KUKA Robocoaster and assessed by a group of 17 test-subjects. The results of the subjective assessment were compared with the numerical metrics. Furthermore, in the thesis an auto-tuning process based on the mean-variance mapping optimization (MVMO) and numerical perception scores was developed by which one is able to automatically tune the parameters to obtain a high well-tuned index, reducing drastically the current high manual tuning times. It was observed that, for a serial robot, the circular motion of the cabin can be well-compensated by the pitch angle. Among the MCAs, the optimal tracking algorithm (ZyRo) and the model predictive control (MPC*) can simulate large amplitude input signals while keeping a high value of the well-tuned index. Furthermore, these algorithms exploit the simulator’s workspace better than the other MCAs and are easily tuned. The ZyRo algorithm produces similar results as the MPC* algorithm, but requires less computational time. The responses of all the MCAs included in this study are similar when using a same low scaling factor (k = 0.4). However, if a larger scale factor is used, the responses of the MCAs change significantly. Finally, it was concluded that a good correlation between the average subjective scores and the objective measures can be achieved. However, due to the large variability of the individual scores, further research is needed to better understand the subjective ratings of simulation rides.