Data-Driven Crowd Simulation with Generative Adversarial Networks
This provides a data-driven solution for crowd simulation in fields like urban planning or gaming, though it is incremental as it builds on existing GAN methods for trajectory generation.
The paper tackles the problem of generating realistic pedestrian crowd simulations by using Generative Adversarial Networks (GANs) to learn from observed trajectories and produce new ones with similar statistical properties, enabling real-time interactions and preserving input characteristics as validated with real-world data.
This paper presents a novel data-driven crowd simulation method that can mimic the observed traffic of pedestrians in a given environment. Given a set of observed trajectories, we use a recent form of neural networks, Generative Adversarial Networks (GANs), to learn the properties of this set and generate new trajectories with similar properties. We define a way for simulated pedestrians (agents) to follow such a trajectory while handling local collision avoidance. As such, the system can generate a crowd that behaves similarly to observations, while still enabling real-time interactions between agents. Via experiments with real-world data, we show that our simulated trajectories preserve the statistical properties of their input. Our method simulates crowds in real time that resemble existing crowds, while also allowing insertion of extra agents, combination with other simulation methods, and user interaction.