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.

A Local Behavior Model for Small Pedestrian Groups

Pedestrians’ behavior in emergency evacuation: Modeling and simulation*

Visually-Guided Pedestrian Crowd Simulation

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

Modifications to Social Force Model

Modeling Dynamic Groups for Agent-Based Pedestrian Crowd Simulations

Understanding and Modeling Collision Avoidance Behavior for Realistic Crowd Simulation

Inserting Virtual Pedestrians into Pedestrian Groups Video with Behavior Consistency

Group Modeling: A Unified Velocity-Based Approach.

A collective motion model based on two-layer relationship mechanism for bi-direction pedestrian flow simulation

Behavior-based Cellular Automaton Model for Pedestrian Dynamics

Relationship Between Behavior Aggressiveness and Pedestrian Dynamics Using Behavior-Based Cellular Automata Model.

Analysis of Cooperation Behaviors and Crowd Dynamics during Pedestrian Evacuation with Group Existence

A Mixed Crowd Movement Model Incorporating Chasing Behavior

Social aspects of collision avoidance: A detailed analysis of two-person groups and individual pedestrians

A Simple and Realistic Pedestrian Model for Crowd Simulation and Application

Modeling Collision Avoidance Behavior With Zero-Speed Pedestrians

Modeling walking behavior of pedestrian groups with floor field cellular automaton approach

Modeling the Mutual Anticipation in Human Crowds with Attention Distractions

Crowd Dynamics of Self-Propelled Individuals with Collision Avoidance Considering Anticipation and Intrusion Aversion

The Walking Behaviour of Pedestrian Social Groups and Its Impact on Crowd Dynamics