Populate Your Game Scene

Abstract

Abstract. We describe a method for populating large virtual environments with crowds of virtual pedestrians. Pedestrians are distributed in the environment by giving them goal destinations to reach. In order to facilitate this population setup, we assign identical destinations to batch of pedestrians, resulting in navigation flows. In order to preserve individuality of pedestrians, navigation is planned with variety and each pedestrians follows its own unique trajectory. This specific navigation planning technique answers user's queries very rapidly. As a result, impact of the insertion of a new navigation flow in the environment is directly observable, and interactive design is achieved. Once the problem of design solved, we address the problem of simulating the virtual population navigation. A level-of-details strategy is used to achieve on-line real-time simulation of large crowds, formed by up to tens of thousands of pedestrians. We illustrate our method on the typical example of a virtual city, and discuss our approach.

Within-crowd immersive evaluation of collision avoidance behaviors

“…Morini et al [Morini et al 2007] and Yersin et al [Yersin et al 2008] proposed a Scalable Motion Planning (SMP) and collision avoidance for a largely crowded scene. We rely on the same spatial scene discretization and highlevel path planification as this approach [Pettré et al 2006] [Pettré 2008]. Van Some previous researches also exploited the visual plausibility of crowds.…”

“…Therefore, instead of computing a trajectory path from scratch, we exploit the possibility of using captured trajectories which includes those missing small scale details. In order to generate various real trajectory path in a given scene model, we defined the navigation graph (Figure 2-a) [Pettré et al 2006] [Pettré 2008] in the walkable area of the scene. Path variants (Figure 2 In the improved model, we increased the number of Real Trajectory Segments (RTS) by reducing rectification thresholds and by mirroring each segment.…”

Long Term Real Trajectory Reuse through Region Goal Satisfaction

“…Our own approach relies on the one described in [19] for the stage of general trajectory planning producing a set of variant paths for large group of pedestrians between two regions in a virtual environment. We differ in the way the variant paths are exploited to produce the individual pedestrian trajectories.…”

“…Each trajectory maintains the information of position, direction and speed of each frame. The first problem we want to address is to fit the longest possible trajectory segment within a set of region-to-region path variants precomputed according to the approach described in [19]. The path variants are constructed from a navigation graph where nodes are (static) collision-free circular regions called vertices (Fig 3 left).…”

“…The Navigation graph[19] illustrated in this city scene (top view) guarantees that no collision with static elements of the environment can occur within the sampled circular regions (left); path variants are automatically built to link distant regions through a list of vertices (right)…”