In the context of crowd simulation, there are a diverse set of algorithms that model steering, the ability of an agent to navigate between spatial locations, while avoiding static and dynamic obstacles. The performance of steering approaches, both in terms of quality of results and computational efficiency, depends on internal parameters that are manually tuned to satisfy application-specific requirements. This paper investigates the effect that these parameters have on an algorithm's performance. Using three representative crowd simulators and a set of established performance criteria, we perform a number of large scale optimization experiments that optimize an algorithm's parameters for a range of objectives. Our experiments show that parameter fitting has a significant impact on the visual fidelity of the simulations produced, the crowd's localized and macroscopic behaviours, as well as the computational efficiency of the steering algorithm. For example, our method automatically finds optimal parameters to minimize turbulence at bottlenecks, reduce building evacuation times, produce emergent patterns, and increase the computational efficiency of an algorithm, in one case by a factor of two. Our study includes an in-depth statistical analysis of the correlations between algorithmic parameters, and performance criteria. The proposed methodology can be applied to any steering algorithm using any set of performance criteria. To our knowledge, this is the first attempt at studying the relationship between a steering algorithm's parameters and its performance, based on parameter fitting.

Download paper in PDF format.

The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.