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.

Collision Avoidance in an Agent System for Crowd Simulation

Understanding and Modeling Collision Avoidance Behavior for Realistic Crowd Simulation

Visually-Guided Pedestrian Crowd Simulation

Flock Based Real Time Collision Avoidance Crowd Simulation

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

Multi-agent Crowd Collision and Path Following Simulation

A new collision avoidance model with random batch resolution strategy

Real-time Control of Individual Agents for Crowd Simulation

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

Adaptive movement behavior for real-time crowd simulation

The crowd cooperation approach for formation maintenance and collision avoidance using multi-agent deep reinforcement learning

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

Crowd evacuation simulation based on hierarchical agent model and physics‐based character control

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

Realtime Collision Avoidance for Mobile Robots in Dense Crowds using Implicit Multi-sensor Fusion and Deep Reinforcement Learning

Crowd Intelligent Cooperative Obstacle Avoidance Method Inspired by Brain Attention Mechanism

Crowd-Aware Robot Navigation for Pedestrians with Multiple Collision Avoidance Strategies via Map-based Deep Reinforcement Learning

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

A Local Behavior Model for Small Pedestrian Groups

Congestion-aware Multi-agent Trajectory Prediction for Collision Avoidance.