The real-time simulation of human crowds has many applications. In a typical crowd simulation, each person ('agent') in the crowd moves towards a goal while adhering to local constraints. Many algorithms exist for specific local 'steering' tasks such as collision avoidance or group behavior. However, these do not easily extend to completely new types of behavior, such as circling around another agent or hiding behind an obstacle. They also tend to focus purely on an agent's velocity without explicitly controlling its orientation. This paper presents a novel sketch-based method for modelling and simulating many steering behaviors for agents in a crowd. Central to this is the concept of an interaction field (IF): a vector field that describes the velocities or orientations that agents should use around a given 'source' agent or obstacle. An IF can also change dynamically according to parameters, such as the walking speed of the source agent. IFs can be easily combined with other aspects of crowd simulation, such as collision avoidance. Using an implementation of IFs in a real-time crowd simulation framework, we demonstrate the capabilities of IFs in various scenarios. This includes game-like scenarios where the crowd responds to a user-controlled avatar. We also present an interactive tool that computes an IF based on input sketches. This IF editor lets users intuitively and quickly design new types of behavior, without the need for programming extra behavioral rules. We thoroughly evaluate the efficacy of the IF editor through a user study, which demonstrates that our method enables non-expert users to easily enrich any agent-based crowd simulation with new agent interactions.

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

In highly dense crowds of humans, collisions between people occur often. It is common to simulate such a crowd as one fluid-like entity (macroscopic), and not as a set of individuals (microscopic, agent-based). Agent-based simulations are preferred for lower densities because they preserve the properties of individual people. However, their collision handling is too simplistic for extreme-density crowds. Therefore, neither paradigm is ideal for all possible densities. In this paper, we combine agent-based crowd simulation with Smoothed Particle Hydrodynamics (SPH), a particle-based method that is popular for fluid simulation. We integrate SPH into the crowd simulation loop by treating each agent as a fluid particle. The forces of SPH (for pressure and viscosity) then augment the usual navigation behavior and contact forces per agent. We extend the standard SPH model with a dynamic rest density per particle, which intuitively controls the crowd density that an agent is willing to accept. We also present a simple way to let agents blend between individual navigation and fluid-like interactions depending on the SPH density. Experiments show that SPH improves agent-based simulation in several ways: better stability at high densities, more intuitive control over the crowd density, and easier replication of wave-propagation effects. Also, density-based blending between collision avoidance and SPH improves the simulation of mixed-density scenarios. Our implementation can simulate tens of thousands of agents in real-time. As such, this work successfully prepares the agent-based paradigm for crowd simulation at all densities. (c) 2021 Elsevier Ltd. All rights reserved.

To simulate the low-level ('microscopic') behavior of human crowds, a local navigation algorithm computes how a single person ('agent') should move based on its surroundings. Many algorithms for this purpose have been proposed, each using different principles and implementation details that are difficult to compare. This paper presents a novel framework that describes local agent navigation generically as optimizing a cost function in a velocity space. We show that many state-of-the-art algorithms can be translated to this framework, by combining a particular cost function with a particular optimization method. As such, we can reproduce many types of local algorithms using a single general principle. Our implementation of this framework, named umans (Unified Microscopic Agent Navigation Simulator), is freely available online. This software enables easy experimentation with different algorithms and parameters. We expect that our work will help understand the true differences between navigation methods, enable honest comparisons between them, simplify the development of new local algorithms, make techniques available to other communities, and stimulate further research on crowd simulation.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Interaction Fields: Sketching Collective Behaviours Adèle Colas, Wouter van Toll, Ludovic Hoyet, Claudio Pacchierotti, Marc Christie, Katja Zibrek, Anne-Hélène Olivier, Julien Pettré

A navigation mesh is a representation of a 2D or 3D virtual environment that enables path planning and crowd simulation for walking characters. Various state-of-the-art navigation meshes exist, but there is no standardized way of evaluating or comparing them. Each implementation is in a different state of maturity, has been tested on different hardware, uses different example environments, and may have been designed with a different application in mind. In this paper, we develop and use a framework for comparing navigation meshes. First, we give general definitions of 2D and 3D environments and navigation meshes. Second, we propose theoretical properties by which navigation meshes can be classified. Third, we introduce metrics by which the quality of a navigation mesh implementation can be measured objectively. Fourth, we use these properties and metrics to compare various state-of-the-art navigation meshes in a range of 2D and 3D environments. Finally, we analyze our results to identify important topics for future research on navigation meshes. We expect that this work will set a new standard for the evaluation of navigation meshes, that it will help developers choose an appropriate navigation mesh for their application, and that it will steer future research in interesting directions. (C) 2020 Elsevier Ltd. All rights reserved.

We present a novel topology-driven method for enhancing the navigation behavior of agents in virtual environments and crowds. In agent-based crowd simulations, each agent combines multiple navigation algorithms for path planning, collision avoidance, and more. This may lead to undesired motion whenever the algorithms disagree on how an agent should pass an obstacle or another agent. In this paper, we argue that all navigation algorithms yield a strategy: a set of decisions to pass obstacles and agents along the left or right. We show how to extract such a strategy from a (global) path and from a (local) velocity. Next, we propose a general way for an agent to resolve conflicts between the strategies of its algorithms. For example, an agent may re-plan its global path when collision avoidance suggests a detour. As such, we bridge conceptual gaps between algorithms, and we synchronize their results in a fundamentally new way. Experiments with an example implementation show that our strategy concept can improve the behavior of agents while preserving real-time performance. It can be applied to many agent-based simulations, regardless of their specific navigation algorithms. The concept is also suitable for explicitly sending agents in particular directions, e.g. to simulate signage. (C) 2020 Elsevier Ltd. All rights reserved.

In highly dense crowds of humans, collisions between people occur often. It is common to simulate such a crowd as one fluid-like entity (macroscopic), and not as a set of individuals (microscopic, agent-based). Agent-based simulations are preferred for lower densities because they preserve the properties of individual people. However, their collision handling is too simplistic for extreme-density crowds. Therefore, neither paradigm is ideal for all possible densities. In this paper, we combine agent-based crowd simulation with the concept of Smoothed Particle Hydrodynamics (SPH), a particle-based method that is popular for fluid simulation. Our combination augments the usual agent-collision handling with fluid dynamics when the crowd density is sufficiently high. A novel component of our method is a dynamic rest density per agent, which intuitively controls the crowd density that an agent is willing to accept. Experiments show that SPH improves agent-based simulation in several ways: better stability at high densities, more intuitive control over the crowd density, and easier replication of wave-propagation effects. Our implementation can simulate tens of thousands of agents in real-time. As such, this work successfully prepares the agent-based paradigm for crowd simulation at all densities.

This paper presents a novel data-driven crowd simulation method that can mimic the observed traffic of pedestrians in a given environment. Given a set of observed trajectories, we use a recent form of neural networks, Generative Adversarial Networks (GANs), to learn the properties of this set and generate new trajectories with similar properties. We define a way for simulated pedestrians (agents) to follow such a trajectory while handling local collision avoidance. As such, the system can generate a crowd that behaves similarly to observations, while still enabling real-time interactions between agents. Via experiments with real-world data, we show that our simulated trajectories preserve the statistical properties of their input. Our method simulates crowds in real time that resemble existing crowds, while also allowing insertion of extra agents, combination with other simulation methods, and user interaction.

We present a novel topology-driven method for improving the navigation of agents in virtual environments. In agent-based crowd simulations, the combination of global path planning and local collision avoidance can cause conflicts and undesired motion. These conflicts are related to the decisions to pass obstacles or agents on certain sides. In this paper, we define an agent’s navigation behavior as a topological strategy amidst obstacles and other agents. We show how to extract such a strategy from a global path and from a local velocity. Next, we propose a simulation framework that computes these strategies for path planning, path following, and collision avoidance. By detecting conflicts between strategies, we can decide reliably when and how an agent should re-plan an alternative path. As such, this work bridges a long-existing gap between global and local planning. Experiments show that our method can improve the behavior of agents while preserving real-time performance. It can be applied to many agent-based simulations, regardless of their specific navigation algorithms. The strategy concept is also suitable for explicitly sending agents in particular directions.

Path planning for walking characters in complicated virtual environments is a fundamental task in simulations and games. A navigation mesh is a data structure that allows efficient path planning. The Explicit Corridor Map (ECM) is a navigation mesh based on the medial axis. It enables path planning for disk-shaped characters of any radius. In this article, we formally extend the medial axis (and therefore the ECM) to 3D environments in which characters are constrained to walkable surfaces. Typical examples of such environments are multi-storey buildings, train stations, and sports stadiums. We give improved definitions of a walkable environment (WE: a description of walkable surfaces in 3D) and a multi-layered environment (MLE: a subdivision of a WE into connected layers). We define the medial axis of such environments based on projected distances on the ground plane. For an MLE with n boundary vertices and k connections, we show that the medial axis has size O(n), and we present an improved algorithm that constructs the medial axis in O(n logn log k) time. The medial axis can be annotated with nearest-obstacle information to obtain the ECM navigation mesh. Our implementations show that the ECM can be computed efficiently for large 2D and multi-layered environments and that it can be used to compute paths within milliseconds. This enables simulations of large virtual crowds of heterogeneous characters in real-time.

Path planning for walking characters in complicated virtual environments is a fundamental task in simulations and games. In this paper, we present an improved definition of the Explicit Corridor Map (ECM), a navigation mesh that allows efficient path planning and crowd simulation for disk-shaped characters of any radius. The ECM is a medial axis (MA) annotated with nearestobstacle information. For a planar environment with n obstacle vertices, the ECM has size O(n) and can be computed in O(n logn) time. We also introduce multi-layered environments (MLEs), in which multiple planar layers are connected by line segment connections. Typical real-world examples are multi-storey buildings, train stations, and sports stadiums. We define the MA and the ECM for multi-layered environments, based on projected distances on the ground plane. For an MLE with n obstacle points and k connections, the MA has size O(n). We present an improved algorithm that constructs the MA and ECM in O(n logn log k) time. Our implementations show that the ECM can be computed efficiently for large 2D and multilayered environments, and that it can be used to compute paths within milliseconds. This enables simulations of large virtual crowds of heterogeneous characters in real-time.

We describe and demonstrate the Explicit Corridor Map (ECM), a navigation mesh for path planning and crowd simulation in virtual environments. For a bounded 2D environment with polygonal obstacles, the ECM is the medial axis of the free space annotated with nearest-obstacle information. It can be used to compute short and smooth paths for disk-shaped characters of any radius. It is also well-defined for multi-layered 3D environments that consist of connected planar layers. We highlight various operations on the ECM, such as dynamic updates, visibility queries, and the computation of paths (indicative routes). We have implemented the ECM as the basis of a real-time crowd simulation framework with path following and collision avoidance. Our implementation has been successfully used to simulate real-life events involving large crowds of heterogeneous characters. The enclosed demo application displays various features of our software.

A navigation mesh is a representation of a 2D or 3D virtual environment that enables path planning and crowd simulation for walking characters. Various state-of-the-art navigation meshes exist, but there is no standardized way of evaluating or comparing them. Each implementation is in a different state of maturity, has been tested on different hardware, uses different example environments, and may have been designed with a different application in mind. In this paper, we conduct the first comparative study of navigation meshes. First, we give general definitions of 2D and 3D environments and navigation meshes. Second, we propose theoretical properties by which navigation meshes can be classified. Third, we introduce metrics by which the quality of a navigation mesh implementation can be measured objectively. Finally, we use these metrics to compare various state-of-the-art navigation meshes in a range of 2D and 3D environments. We expect that this work will set a new standard for the evaluation of navigation meshes, that it will help developers choose an appropriate navigation mesh for their application, and that it will steer future research on navigation meshes in interesting directions.

Modern simulations feature crowds of AI-controlled agents moving through dynamic environments, with obstacles appearing or disappearing at run-time. A dynamic navigation mesh can represent the traversable space of such environments. The A* algorithm computes optimal paths through the dual graph of this mesh. When an obstacle is inserted or deleted, the mesh changes and agents should re-plan their paths. Many existing re-planning algorithms are too memory-intensive for crowds, or they cannot easily be used on graphs where vertices and edges are added or removed. In this paper, we present Dynamically Pruned A* (DPA*), an extension of A* for re-planning optimal paths in dynamic navigation meshes. DPA* has similarities to adaptive algorithms that make the A* heuristic more informed based on previous queries. However, DPA* prunes the search using only the previous path and its relation to the dynamic event. We describe the four re-planning scenarios that can occur; DPA* uses different rules in each scenario. Our algorithm is memory-friendly and robust against structural changes in the graph, which makes it suitable for crowds in dynamic navigation meshes. Experiments show that DPA* performs particularly well in complex environments and when the dynamic event is visible to the agent. We integrate the algorithm into crowd simulation software to model large crowds in dynamic environments in real-time.

This paper presents a GPU-accelerated approach for improving the approximated construction of Generalized Voronoi Diagrams (GVDs). Previous work has shown how to render a GVD onto the GPU framebuffer, and copy it to the CPU for extraction of a high-quality diagram. We improve upon this technique by performing more computations in parallel on the GPU, and reducing the amount of data transferred to the CPU. We also design a multi-tiled construction technique that overcomes hardware limitations and enables much higher rendering resolutions, thus reducing discretization errors. Next, we extend our approach to create an Explicit Corridor Map navigation mesh, which is an efficient data structure for path planning in modern crowd simulation systems. The new implementation allows much faster construction of GVDs and navigation meshes at virtually infinite resolutions.