Figure 1: A scenario where 25 agents (in orange) move to the right and 25 other agents (in purple) move down. This figure shows the crowd after 10 seconds, simulated using various algorithms in the umans framework. From left to right: RVO [van den

Fig. 1. Our objective is to understand whether and to what extent providing haptic rendering of collisions during navigation through a virtual crowd (right) makes users behave more realistically. Whenever a collision occurs (center), armbands worn on the arms locally vibrate to render this contact (left). We carried out an experiment with 23 participants, testing both subjective and objective metrics regarding the users' path planning, body motion, kinetic energy, presence, and embodiment.

Figure 1: Waving case from left to right: two Intelligent Virtual Agents (IVA) -A and B -observe another one; visual motion features -body orientation, gesture amplitude -are computed on the observed IVA's motions from each viewpoint; B reacts

Figure 1: Our topology-aware camera control system works as follows: starting from a virtual environment with its navigation mesh in blue (left), a collection of camera tracks are generated by clustering points obtained via ray casts (green) generated from a topological skeleton representation of the navigation mesh (center-left).The camera is then controlled in real-time by a physical system that follows a target on the best camera track in order to film an actor navigating in the environment (center-right) and the final result (right).