Anne-Hélène Olivier

Experiment-based modeling, simulation and validation of interactions between virtual walkers

An interaction occurs between two humans when they walk with converging trajectories. They need to adapt their motion in order to avoid and cross one another at respectful distance. This paper presents a model for solving interactions between virtual humans. The proposed model is elaborated from experimental interactions data. We first focus our study on the pair-interaction case. In a second stage, we extend our approach to the multiple interactions case. Our experimental data allow us to state the conditions for interactions to occur between walkers, as well as each ones role during interaction and the strategies walkers set to adapt their motion. The low number of parameters of the proposed model enables its automatic calibration from available experimental data. We validate our approach by comparing simulated trajectories with real ones. We also provide comparison with previous solutions. We finally discuss the ability of our model to be extended to complex situations.

A step-by-step modeling, analysis and annotation of locomotion

Annotating unlabeled motion captures plays an important role in Computer Animation for motion analysis and motion edition purposes. Locomotion is a difficult case study as all the limbs of the human body are involved whereas a low‐dimensional global motion is performed. The oscillatory nature of the locomotion makes difficult the distinction between straight steps and turning ones, especially for subtle orientation changes. In this paper we propose to geometrically model the center of mass trajectory during locomotion as a C‐continuous circular arcs sequence. Our model accurately analyzes the global motion into the velocity‐curvature space. An experimental study demonstrates that an invariant law links curvature and velocity during straight walk. We finally illustrate how the resulting law can be used for annotation purposes: any unlabeled motion captured walk can be transformed into an annotated sequence of straight and turning steps. Several examples demonstrate the robustness of our approach and give comparison with classical threshold‐based techniques. Copyright

Cave size matters: Effects of screen distance and parallax on distance estimation in large immersive display setups

When walking within a CAVE-like system, accommodation distance, parallax, and angular resolution vary according to the distance between the user and the projection walls, which can alter spatial perception. As these systems get bigger, there is a need to assess the main factors influencing spatial perception in order to better design immersive projection systems and virtual reality applications. In this paper, we present two experiments that analyze distance perception when considering the distance toward the projection screens and parallax as main factors. Both experiments were conducted in a large immersive projection system with up to 10-meter interaction space. The first experiment showed that both the screen distance and parallax have a strong asymmetric effect on distance judgments. We observed increased underestimation for positive parallax conditions and slight distance overestimation for negative and zero parallax conditions. The second experiment further analyzed the factors contributing to these effects and confirmed the observed effects of the first experiment with a high-resolution projection setup providing twice the angular resolution and improved accommodative stimuli. In conclusion, our results suggest that space is the most important characteristic for distance perception, optimally requiring about 6- to 7-meter distance around the user, and virtual objects with high demands on accurate spatial perception should be displayed at zero or negative parallax.

Biomechanics of Walking in Real World: Naturalness we Wish to Reach in Virtual Reality

In most virtual reality (VR) simulations the virtual world is larger than the real walking workspace. The workspace is often bounded by the tracking area or the display devices. Hence, many researchers have proposed technical solutions to make people walk through large virtual spaces using various types of metaphors and multisensory feedback. To achieve this goal it is necessary to understand how people walk in real life. This chapter reports biomechanical data describing human walking including kinematics, dynamics and energetics knowledge for straight line and nonlinear walking. Reference and normative values are provided for most of these variables, which could help developers and researchers improve the naturalness of walking in large virtual environments, or to propose evaluation metrics. For each section of this chapter, we will provide some potential applications in VR. On the one hand, this type of knowledge could be used to design more natural interaction devices such as omnidirectional treadmills, walk-in-place methods, or other facilities. A specific section is dedicated to comparisons between treadmill and ground walking as it is one of the most popular approaches in VR. On the other hand, this knowledge could also be useful to improve the quality of multisensory feedback when walking, such as adding sounds, vibrations, or more natural camera control.

A Velocity-Curvature Space Approach for Walking Motions Analysis

This paper presents an automatic turns detection and annotation technique which works from unlabeled captured locomotion. Motion annotation is required by several motion capture editing techniques. Detection of turns is made difficult because of the oscillatory nature of the human locomotion. Our contribution is to address this problem by analyzing the trajectory of the center of mass of the human body into a velocity-curvature space representation. Our approach is based on experimental observations of carefully captured human motions. We demonstrate the efficiency and the accuracy of our approach.

Curvature–velocity analysis to identify turning steps while walking

Non-straight steps represent up to 35–45% of the total amount of steps in daily walking (Glaister et al. 2007). This behaviour is then a fundamental motor activity that allows collision avoidance or direction changing. Curved walking was recently investigated to explore dynamic stability (Segal et al. 2008), head anticipation (Prévost et al. 2003), trajectory generation (Hicheur et al. 2007) or feet placement strategy (Hase and Stein 1999). Two main feet strategies were identified, according to the stance foot placed in front for braking: the ‘spin turn’ and the ‘step turn’. The difference between the two strategies depends on which foot is the stance one while turning: the inside one (spin) or the outside one (step). This definition assumes that the turn is performed in one step or at least that there is one preponderant step. This question remains unsolved. Nonetheless, several procedures were applied to analyse curved paths. Glaister et al. (2007) used manual video analysis to classify each step of a trajectory. Other authors took interest in the direction of progression to identify a medio-lateral change in a subject’s trajectory compared to a straight path (Vallis and McFadyen 2003). Trajectory could also be totally constrained: a subject has to follow a predefined path drawn on the floor (Fuller et al. 2007). Therefore, the aim of the present work is to provide an automatic method that first detects turning steps within a natural trajectory and then identifies the stepping strategy while turning. This method is based on the analysis of the speed/curvature relation while walking.

Local kinematics of human walking along a turn

Human walking along a curved path is a fundamental motor synergy since it allows a subject to navigate in security. Indeed, humans have to turn in a corridor, to avoid obstacles or to encounter somebody which is not on their initial path. Global approaches consider the trajectory of a specific point (centre of mass, CoM; top of the head, etc.) as a whole. A close relation, named power law relation, has been demonstrated between geometry of the path to follow and velocity while walking (Vieilledent et al. 2001). This relation depends on the shape of this repeated predefined trajectory (Hicheur et al. 2005). It was also underlined a stereotypy of both geometry and velocity profile during goal directed walking trajectories (Hicheur et al. 2007). This suggests that the walking trajectory should be planned as a whole. Local approaches consider specific events of the locomotor trajectory. The power law relation between radius of curvature and velocity was explored on the particular point of one 908 turn that matches the minimal radius of curvature (Olivier and Crétual 2007). It did highlight a discrete power law relation for each trials of each subject. This would strengthen the idea of a global planning of the trajectory. Some studies also take interest in foot strikes along a path. Two main strategies for turning were identified, according to the stance foot forward during the initiation of the turn. The ‘spin turn’ is a turn on the left with the left stance foot forward (respectively right and right), and the ‘step turn’ represents a turn on the left with the right stance foot forward (respectively right and left) (Patla et al. 1991; Hase and Stein 1999). These strategies imply different organisations. In the transversal plane, rotations are performed in the opposite direction, namely internal rotations for the spin turn and external rotations for the step turn (Taylor et al. 2005). Muscular demand is then higher for the spin turn. The step turn offers a stable base of support (Patla et al. 1991; Hase and Stein 1999; Taylor et al. 2005). On the contrary, the spin turn threatens stability because the CoM goes outside the base of support. Feet are moreover closer, which facilitates the turning motion by decreasing the moment of inertia but increases falling risks (Taylor et al. 2005). Let us denote that these protocols did not take interest in the preferred strategy to adopt. The geometry of the trajectory influences the global kinematics of curved locomotion (Hicheur et al. 2005). Then, we wondered whether it influences local parameters such as the foot strikes and more precisely the spin and step turn strategies.