Abstract:For walking pedestrians, when they are blocked by obstacles or other pedestrians, they adjust their speeds and directions to avoid colliding with them, which is called collision avoidance behavior. This behavior is the most complex part of pedestrians' walking processes and its modeling and simulation are the keys to realistic crowd simulation, which serves as the foundation for various applications. However, most existing methods either lack the representation power to accurately model the complex collision behavior or do not model it explicitly, which leads to a poor level of realism of the simulation. To realize realistic crowd simulation, we propose to analyze, understand, and model the collision avoidance behavior in a data-driven way. First, to automatically detect collision avoidance behavior for further analysis, we propose a domain transformation algorithm that detects it by transforming the trajectories in the spatial domain into a new domain where the behavior is much more apparent and is thus easier to detect. The new domain also provides a new perspective for understanding collision avoidance behavior. Second, since there are no mature metrics to evaluate the level of realism, we propose a new evaluation metric based on the least-effort theory, which evaluates the realism of collision avoidance behavior by its physical and mental consumption. This evaluation metric also provides the foundation of modeling. Third, for realistic crowd simulation, we design a reinforcement learning model. It trains agents with our proposed reward function that models pedestrians' intrinsic needs of "reducing effort consumption'' and thus can guide agents to behave realistically when avoiding collisions. Extensive experiments show our model is 55.9% and 52.5% more realistic in collision avoidance behavior than the best baselines on two real-world datasets. We release our codes at https://github.com/tsinghua-fib-lab/TECRL.

Modeling Gap Seeking Behaviors for Agent-based Crowd Simulation

Modifications to Social Force Model

Learning Behavior Patterns from Video for Agent-Based Crowd Modeling and Simulation

Group Modeling: A Unified Velocity-Based Approach.

Understanding and Modeling Collision Avoidance Behavior for Realistic Crowd Simulation

Modeling Dynamic Groups for Agent-Based Pedestrian Crowd Simulations

Model-based Crowd Behaviours in Human-solution Space

Learning Behavior Patterns from Video: A Data-driven Framework for Agent-based Crowd Modeling.

A Group Behavior Model Of Real-Time Crowds In Virtual Environments

Data-Driven and Collision-Free Hybrid Crowd Simulation Model for Real Scenario

Collision Avoidance in an Agent System for Crowd Simulation

A Hyper-Heuristic Framework for Agent-Based Crowd Modeling and Simulation: (extended Abstract).

Agent-based simulation of collective cooperation: from experiment to model

A Hyper-Heuristic Framework for Agent-Based Crowd Modeling and Simulation

Continuous agent-based modeling of adult-child pairs based on a pseudo-energy: Relevance for public safety and egress efficiency

Crowd Behavior Simulation Model Based on Information-Gap Theory

Visually-Guided Pedestrian Crowd Simulation

Real-time Control of Individual Agents for Crowd Simulation

Simulation Model of Self-Organizing Pedestrian Movement Considering Following Behavior.

A Force-Based Model for Adaptively Controlling the Spatial Configuration of Pedestrian Subgroups at Non-Extreme Densities