# Data-Driven Crowd Simulation with Generative Adversarial Networks

## TL;DR

This paper introduces a GAN-based data-driven crowd simulation method that learns from observed pedestrian trajectories to generate realistic, real-time crowd behaviors with statistical fidelity and interactive capabilities.

## Contribution

It presents a novel GAN-based approach for learning and generating pedestrian trajectories that mimic real-world crowd patterns in real time.

## Key findings

- Simulated trajectories preserve statistical properties of real data
- The system enables real-time crowd simulation with user interaction
- Allows insertion of extra agents and integration with other methods

## Abstract

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.