Algorithms for Microscopic Crowd Simulation: Advancements in the 2010s

The real-time simulation of human crowds has many applications. Simulating how the people in a crowd move through an environment is active and ever-growing research topic. Most focuses on microscopic (or ‘agent-based’) crowd-simulation methods that model behavior each individual person, from which collective can then emerge. This state-of-the-art report analyzes advanced since year 2010. We focus most popular area within paradigm, local navigation, notably collision avoidance between agents. discuss four categories algorithms this (force-based, velocity-based, vision-based, data-driven) have either emerged or grown last decade. also analyze conceptual computational (dis)advantages category. Next, we extend discussion to other types navigation (such as group combination with path planning), review work evaluating quality simulations. Based observed advancements 2010s, conclude by predicting will evolve future. Overall, expect significant growth data-driven learning-based agent increasing number re-group multiple ‘levels’ into one principle. Furthermore, observe clear need for new ways (real simulated) behavior, important quantifying realism choosing right at time.