Section: New Results


Participants : Julien Pettré [contact] , Anne-Hélène Olivier, Julien Bruneau, Jonathan Perrinet, Kevin Jordao, David Wolinski.

Analysis of Locomotion Trajectories during Collision Avoidance

The experimental observation of physical interactions between real walkers is for us a great source of inspiration for the design of realistic microscopic models of crowd simulation. This year, we have continued analysing locomotion trajectories of real walkers during collision avoidance tasks. Analysis focused on individual strategies and role set to solve such a reciprocal interaction. Our analysis revealed that walkers combine re-orientation and speed adaptations to avoid collisions, but more importantly, that the strategies, as well as the global amount of adaptations is dependent on the role each one has in the avoidance (e.g., passing first, giving way). Our results are reported in [13] . In addition, we inspected the role of psychologic factors on the metrics of interactions [27] .

Evaluation of Locomotion Trajectories performed in Virtual Reality

Virtual Reality rooms are physically limited in space, and prevent users virtually walking by really walking in larger virtual spaces: a locomotion interface is employed to overcome this issue. The interface is composed of a peripheral device, such as joystick, as well as of a software component which transform users' actions on the peripheral device into a virtual locomotion. In this work, we wondered if users where performing similar trajectories in virtuo than in vivo: such question is important when aiming at using VR form motion analysis purpose. We evaluated the bias introduced by several couples of devices and software components during the execution of goal directed locomotion tasks. As reported in [7] , impressive similarities on the formed trajectories even when the device control motions are radically different in comparison with walking motions.

Virtual Populations for large-scale digital environments and Cultural Heritage Applications

We are developing techniques dedicated to the animation of large virtual populations at very low computational cost based on the crowd patches techniques. Crowd patches can be described as 3D animated textures that small groups animations. They are composed in space to form large population. This year, we coupled the crowd patches approach with mutable shape models: such association enable users cdesigning patches composition in an interactive manner, as introduced in [22] . We applied those techniques to design populations of some old Malaysian trading ports [25] .

Macroscopic derivations of microscopic simulation models

Crowd phenomenon exhibit macroscopic structures which derive from the combination of local interactions between individuals. Together with the IMT in Toulouse in the frame of the ANR-Pedigree project (term. 2012), the microscopic models developed in our team has been derived into macroscopic models to demonstrate their ability to provoke the mergence of some typical macroscopic structures [35] , [36] .