Sketching for Real-time Control of Crowd Simulations

Controlling the behaviour of a crowd simulation typically involves tuning of a system's parameters through trial and error, a time-consuming process relying on knowledge of a potentially complex parameter set. Numerous graphical control approaches have been proposed to allow the user to interact with a simulation intuitively. This research investigates the use of a real-time sketch-based approach for crowd simulation control. This is done by modifying the environment of the simulation. Users can create entrances/exits, barriers and flow lines in real-time on top of an environment. This process requires a data structure to represent the environment and navigate the crowd through it. Two alternatives are presented: grid and navigation mesh. A detailed comparison shows that the navigation mesh is a more scalable approach since it uses less memory, has a similar pathfinding time, and is a better structure to represent the environment than the grid. The thesis also presents extensions to the sketch-based approach in the form of novel control tools, including storyboards to define the journey of the crowd, a timeline interface to simulate events through the day, and a sketch-based group storyboard to link behaviours and paths to be followed by a group. These tools are used to create two complex scenarios to exemplify possible applications of the sketch-based approach. The work on timelines also raises a new problem for an approach that dynamically modifies an environment in real-time which is 'when does the crowd know about the change?' Some initial solutions to how this should be handled are presented. The sketch-based system is evaluated by comparing it to a validated commercial system called MassMotion. The comparison takes into account the plausibility of the simulation and usability of the user interface. A user study is carried out to evaluate the graphical user interface of both systems. Formal evaluation methods are used to make the comparison: the benchmark suite 'steersuite', an adapted version of the Keystroke-Level Model (KLM) and the System Usability Scale (SUS). The results show that the sketch-based approach is faster and easier to use than MassMotion, but with fewer control options. An implementation of the sketching interface in a Virtual Reality environment is also considered. However, when compared to the desktop interface using a proposed adaptation to KLM for VR, the results show that sketching in a VR environment is slower and less accurate than the desktop version.