IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS, VOL. 43, NO. 3, MAY 2013

265

Motion Scaling for High-Performance Driving Simulators Alain Berthoz, Willem Bles, H. H. B¨ulthoff, B. J. Correia Gr´acio, Philippus Feenstra, Nicolas Filliard, R. H¨uhne, Andras Kemeny, M. Mayrhofer, M. Mulder, H. G. Nusseck, P. Pretto, Gilles Reymond, R. Schl¨usselberger, J. Schwandtner, H. Teufel, Benjamin Vailleau, M. M. (Ren´e) van Paassen, Manuel Vidal, and Mark Wentink

Abstract—Advanced driving simulators aim at rendering the motion of a vehicle with maximum fidelity, which requires increased mechanical travel, size, and cost of the system. Motion cueing algorithms reduce the motion envelope by taking advantage of limitations in human motion perception, and the most commonly employed method is just to scale down the physical motion. However, little is known on the effects of motion scaling on motion perception and on actual driving performance. This paper presents the results of a European collaborative project, which explored different motion scale factors in a slalom driving task. Three state-of-the-art simulator systems were used, which were capable of generating displacements of several meters. The results of four comparable driving experiments, which were obtained with a total of 65 participants, indicate a preference for motion scale factors below 1, within a wide range of acceptable values (0.4–0.75). Very reduced or absent motion cues significantly degrade driving performance. Applications of this research are discussed for the design of motion systems and cueing algorithms for driving simulation. Index Terms—Human factors, road vehicles, virtual reality.

Manuscript received April 12, 2010; revised February 29, 2012; accepted December 3, 2012. Date of current version April 15, 2013. This work was supported in part by the Motion cueing for Vehicle Simulators (MOVES) EUREKA#3601 European Research Project, which aims at increasing the scientific knowledge on the human multi-sensory perception of motion in virtual environments, and to explicitly define the possibilities and limitations of several high-end European driving simulators, and by the World Class University program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology under Grant R31-10008. This paper was recommended by Associate Editor Y. Lin of the former IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans (2011 Impact Factor: 2.123). A. Berthoz and M. Vidal are with the Laboratory of Physiology of Perception and Action, CNRS et Coll`ege de France, 94736 Paris, France (e-mail: [email protected]; [email protected]). W. Bles, P. Feenstra, and M. Wentink are with TNO-Human Factors Research Institute, 3769 ZG Soesterberg, The Netherlands (e-mail: [email protected]; [email protected]; [email protected]). H. H. B¨ulthoff is with Max Planck Institute-Biological Cybernetics, 72076 T¨ubingen, Germany, and also with the Department of Brain and Cognitive Engineering, Korea University, Seoul 136-713, Korea. B. J. Correia Gr´acio, M. Mulder, and M. M. van Paassen are with the Delft University of Technology, 2628 CN Delft, The Netherlands (e-mail: [email protected]). N. Filliard, A. Kemeny, G. Reymond, and B. Vailleau are with RENAULT, 78288 Guyancourt, France (e-mail: [email protected]; [email protected]; [email protected]; benjamin.vailleau@renault. com). R. H¨uhne, M. Mayrhofer, R. Schl¨usselberger, and J. Schwandtner are with AMST-Systemtechnik GmbH, 5282 Ranshofen, Austria. H. G. Nusseck, P. Pretto, and H. Teufel are with Max Planck InstituteBiological Cybernetics, 72076 T¨ubingen, Germany. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TSMC.2013.2242885

I. INTRODUCTION HEN driving a car, it is generally admitted that visual information is the primary sensory feedback used by drivers to keep their car on the road. This statement may be overrated [1], however, as it occludes the multisensory nature of human control. More specifically, a growing body of research has pointed out that inertial sensory cues are also taken into account while driving [2]–[6]. When simulating forward travel with visual cues only, significant perception biases can appear, such as speed underestimation effects [5], as well as asymmetries in the extrapolation of accelerated and decelerated self-motion [7]. These biases should be taken into account when using fixed-based driving simulators. Normal motion is signaled by the human vestibular and proprioceptive sensory systems, and these inertial cues can be reproduced in moving-based simulators by controlling their mechanical actuators within certain limits in terms of displacement, velocity, and acceleration. While the Gough–Stewart hexapod configuration of linear actuators has become a standard in flight simulation over the years, other designs were explored for car driving simulators to better reproduce the long periods and fast variations of horizontal acceleration which are typical for many driving situations. The use of linear rails with horizontal displacements of several meters is now a common solution, since the development of the first high-performance driving simulators [8], [9], but actuator size can be a critical economical issue. Besides, should the simulator reproduce the actual motion of the car, or the drivers’ sensation of the car motion? Several characteristic limitations of the human perception of self-motion can be used, to reduce the displacement of the simulator without affecting the subjective realism of the motion cues. Human motion perception models can be integrated into motion cueing algorithms to achieve a trajectory that would still lead to a perception of self-motion as close as possible to the corresponding real situation, yet within the limits of the motion system at hand [10], [11]. However, this strategy requires comprehensive knowledge of human multisensory self-motion perception, which is not fully accessible. There is an ambiguity between gravity and acceleration signals in the vestibular system that can be employed to render apparent sustained linear accelerations by simply tilting the simulator cabin. This technique is referred to as “tilt coordination” in reference to flight simulation [12], but is only effective for slow variations of acceleration [13], and is, therefore, often less effective in car driving simulation [14].

W

2168-2291/$31.00 © 2013 IEEE

266

IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS, VOL. 43, NO. 3, MAY 2013

Another interesting property, which has not been modeled yet, is that the perception of simulated self-motion can tolerate significant discrepancies between the physical and visual motion cues [15]–[17]. In particular, the exact “one-to-one” rendering of acceleration has been shown not to be necessarily the best solution for the perceived coherence of self-motion in simulators [18]–[20]. The origin of this effect is still debated, but the application of a subunity “scale factor” to the simulation of accelerations has been widely accepted in the flight and driving simulation communities as an effective means to decrease the requirements for linear displacement. Yet, the existence of an “optimal” motion scale factor is still questionable. It is also unclear whether different scale factors may be applied to different axes of motion. For instance, when driving a curve under normal conditions, the car heading rate and lateral acceleration are physically coupled; therefore, applying differential scale factors in a simulator may interfere with the drivers’ internal model of the motion of their vehicle. Driving simulators are increasingly used for industrial vehicle design, human factors research, or driver training. The accurate knowledge of motion cueing requirements is a critical issue for the assessment of their validity. In this paper, we investigate the specific effect of scale factors for rendering car motion, by comparing experimental results obtained in three high-performance and innovative driving simulators. The exceptional motion capacities of these recent simulators, and the choice of a common reference slalom driving task, allow for a one-to-one reproduction of accelerations as well as the exploration of motion scaling below and above unity. These results were obtained within the MOVES European research consortium composed of industrial and research partners. This paper is structured as follows. In the first section, we present the configurations of the three simulators; the second section describes four driving experiments conducted on a common scenario; the last section provides a comparison and discussion of the results, which confirm the possibility to use motion gains lower than unity under practical conditions of driving simulation, within a broad range of acceptable gains. II. DRIVING SIMULATORS AND SCENARIO A. DESDEMONA DESDEMONA (see Fig. 1) is a moving-based researchsimulator, which is located at TNO (Soesterberg, The Netherlands), that was designed with a special focus on spatial disorientation demonstrations, flight simulation, and driving simulation. The unique layout (i.e., the unique ability of sustained g-loading and unlimited attitude control) of DESDEMONA is the result of a joint venture of TNO and AMST (Ranshofen, Austria). Although the fundamental design and specification of the platform were defined by the two partners, AMST developed, manufactured, and commissioned the platform. DESDEMONA has been operational since 2006. The simulator has six degrees of freedom (DoF): The cabin is mounted in a gimbaled system (3 DoF, >2π rad), which as a whole can move vertically along a heave axis (1 DoF, ±1 m) and horizontally along a linear arm (1 DoF, ±4 m). This structure then can rotate

Fig. 1.

DESDEMONA motion simulator (TNO/AMST).

Fig. 2.

ULTIMATE driving simulator (RENAULT).

as a whole around a central axis to facilitate centrifugal motion (1 DoF,