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.

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

Visually-Guided Pedestrian Crowd Simulation

A Mixed Crowd Movement Model Incorporating Chasing Behavior

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

Modeling the Mutual Anticipation in Human Crowds with Attention Distractions

Understanding and Modeling Collision Avoidance Behavior for Realistic Crowd Simulation

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

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

Anticipation Dynamics of Pedestrians Based on the Elliptical Social Force Model.

Crowd Dynamics Based on a Modified Social Force Model Considering Predicting Collision Behaviors

A Dual Adaptive Cellular Automaton Model Based on a Composite Field and Pedestrian Heterogeneity

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

Dynamics characteristic of pedestrians' particular overtaking behavior based on an improved social force model

Exploring crowd persistent dynamism from pedestrian crossing perspective: An empirical study

Effects of Crowding Perception on Self-organized Pedestrian Flows Using Adaptive Agent-based Model

High-statistics modeling of complex pedestrian avoidance scenarios

Experimental study of the behavioural mechanisms underlying self-organization in human crowds

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

Modelling Physical Contacts to Evaluate the Individual Risk in a Dense Crowd

Group Modeling: A Unified Velocity-Based Approach.

A modified social force model for crowd evacuation considering collision predicting behaviors