Skip to main content
GameDev.net gamedev.net

Crowd Simulation News

The latest Crowd Simulation coverage curated for game developers.

Boids in Unity: Flocking, Swarming, and Crowds

git-amend shows how to build Craig Reynolds’ Boids in Unity from scratch, covering separation, alignment, cohesion, obstacle avoidance, and boundary steering. The tutorial is aimed at devs who want flocking, …

Video git-amend · 1 month, 3 weeks ago
How Affect Propagates among LLM Agents: Emergent Emotional Contagion in Crowd Simulation

LLM-driven crowd sims are starting to show believable emotional contagion without hard-coded mood transfer. A new multi-agent system lets each agent perceive, appraise, and express affect through an LLM, while …

Research arXiv cs.GR · 2 months ago
LLM-Driven Personalities for Decision Making in Emergency Simulations

A new arXiv paper shows LLMs can steer evacuation-sim agents through personality-driven prompts, not just fixed rules. The interesting bit for game AI is that OCEAN traits produced noticeably different …

Research arXiv cs.GR · 2 months, 4 weeks ago
EvoFlock: evolved inverse design of multi-agent motion

Craig Reynolds is applying inverse design to flocking and other multi-agent motion, using a genetic algorithm to tune behavior toward a user-defined fitness function. For developers, the interesting bit is …

Research arXiv cs.GR · 3 months ago
Improved visual-information-driven model for crowd simulation and its modular application

A crowd-simulation paper from arXiv proposes a data-driven model that uses refined visual cues plus explicit exit signals, and it was tested on four common layouts: bottleneck, corridor, corner, and …

Research arXiv cs.GR · 4 months, 1 week ago
Contrastive Multi-Modal Hypergraph Reasoning for 3D Crowd Mesh Recovery

A new approach to 3D crowd reconstruction leverages multi-modal hypergraph reasoning, addressing challenges like occlusions and depth ambiguity. This method enhances the accuracy of crowd mesh recovery, which is crucial …

Research arXiv cs.GR · 4 months, 2 weeks ago
BEHAVE: Real-Time Modeling of Human Systems as Observable Complex Dynamical Systems and Operational Objects for Physical AI

arXiv cs.GR details BEHAVE, a real-time model that treats interacting people as a persistent HumanSystem rather than isolated tracks. The paper says that framing improves group discrimination on pedestrian data …

Research arXiv cs.GR · 4 months, 2 weeks ago
CrowdVLA: Embodied Vision-Language-Action Agents for Context-Aware Crowd Simulation

CrowdVLA introduces a paradigm shift in crowd simulation by modeling agents as Vision-Language-Action entities. This approach enhances realism by allowing agents to interpret their environment contextually, making decisions based on …

Research arXiv cs.GR · 5 months, 3 weeks ago
Gen-C: Populating Virtual Worlds with Generative Crowds

Gen-C introduces a groundbreaking framework for simulating crowds in virtual environments, leveraging Large Language Models to generate synthetic datasets. This innovation allows developers to create more realistic agent interactions without …

Research arXiv cs.GR · 6 months ago