Using Functional Populations to Create More Realistic Simulations

They are walking with us to work in the morning. They are eating next to us in cafes. They are grabbing some milk as we wander down the aisle of a grocery store. They are tying their child’s shoe as we pass by. They are delivering packages in our office. They are playing soccer in the field below our window. They are rushing to their next meeting as we are returning from one of our own. They are laughing together as we pass on the sidewalk. They are chatting on the phone as wait for the elevator. People inhabit many spaces in our daily lives. Some we interact with. Others are merely a part of the backdrop. Throughout our lives we are conditioned to have expectations of the people around us. These expectations are often predicated by characteristics of the people and of the overall context. When a person’s characteristics or behaviors deviate from our expectations or from those found in other individuals in the context, we notice. These expectations of human behavior carry over to simulations and games. The characters that provide this backdrop of activity are often termed NPCs (Non-Player Characters). NPCs add richness to simulated environments. A world without them would seem barren and artificial. People are often a major part of the setting and back story of a simulation or game. Furthermore, NPCs are more believable if they perform and have a life outside of their scripted interactions with the player or trainee. Most crowd simulators focus on maneuvering large numbers of characters from one location to another without colliding. Some, such as for theatrical battle scenes, also have a limited number of reactive behaviors (e.g. hand to hand combat). What is really desired is much richer. We would like to see a heterogeneous population, not just in terms of clothing or body shape, but also in terms of meaningful behaviors. Furthermore, these behaviors should be appropriate to the context (i.e. time and place) and include meaningful interactions with the objects in the environment and with other characters. We term such populations functional crowds, meaning that the characters are actually inhabiting and functioning within a space instead of merely passing through it. Moreover, we want these simulations to be easy to author and modify such that different scenarios can be explored to determine the most effective for the given application. Finally, it would be great if there were plausible variations in the simulations each time they run so that they do not become stale and ineffective. One avenue is to use character roles to help specify behaviors. When we think about what defines various roles, behavioral expectations play a big part. If we link behaviors with roles and then define our virtual population based on groups of characters with certain roles, then we will have created a heterogeneous population whose characters are already associated with some meaningful behaviors. Next we want to ensure that these behaviors are performed in the right context in terms of the environment setting and script of the simulation. A key here is to capture some of the semantics of the actions and objects in the environment and any ties that they have to each other (e.g. sitting in a chair). When these semantics are captured and parameterized in general terms, they are independent of a single scenario and can be reused, making the overall framework more extensible to new domains. Such a representation can also span the divide between data needed at an animation level (e.g. positions, sites, joint angles, etc) and data more relevant at a planning or authoring level (e.g. action names and their impacts on the virtual world, etc), enabling a game designer or simulation author to focus on the high level aspects of the scenario. Such a representation could also allow non-programmers (e.g. Subject Matter Experts (SMEs)) to author simulations directly, promoting faster prototyping, more experimentation, and more refined scenarios. Finally, different types of actions can be specified to create richer, emergent behaviors. Scheduled actions could arise from the specified roles to help establish context (e.g. students attending a class) or even promote a storyline. Reactive actions are triggered by contextual events and add life to the simulation (e.g. characters waving hello when they see each other). Opportunistic actions are based on fulfilling needs and add realism (e.g. grabbing a cup of coffee when energy is low). Aleatoric actions are random, but structured by choices and distributions. These statistically driven actions further promote plausible variations in a simulation without taxing a scenario author (e.g. working in an office includes varying amounts of typing, reading, talking on a phone, etc). A framework such as the one discussed could be used to set a more reasonable backdrop for training simulations of all sorts. For example, it could be used to established normalcy for a population before introducing aberrant behavior that a trainee should detect and respond to. Furthermore, in data driven systems such as the one outlined, it is possible to dynamically alter the behavior of a character or set of characters (e.g. switch a character’s role from shopkeeper to insurgent or instruct a group of characters to suddenly gather at a location). For more information and examples of preliminary work in this arena, please visit: http://cs.gmu.edu/~gaia/CAROSA/

Using Functional Populations to Create More Realistic Simulations

Upcoming Events

Wildfire Technology Summit

Air Dominance Summit

Next Generation UAS

Armored Vehicles USA Conference

Veterans Healthcare 2026

Counter UAS Summit

Follow Us

Latest Webinars

Responsible and Transparent Approaches to AI in Biometric Algorithm

Securing the Defence Industrial Base: Mitigating Risk and Delivering Resiliency in Physical and Digital Supply Chains

Treatment Option for Two Subsets of Challenging to Treat Major Depressive Disorder in Adults

Recommended

Facial Recognition: Art or Science?

Facial Recognition: Art or Science?

NORAD’s Perspective on Radar: Interview with Lieutenant General Pierre St-Amand...

OBIM's View On Biometrics

Controlling the Waters Biometric Security at Port Everglades

Trump and Tricare: How will the new administration affect Military Healthcare?

Trump and Tricare: How will the new administration affect Military Healthcare?

How Will the New Administration Affect Military Healthcare?

Unmanned Solutions for Ground Combat Systems: with Johns Hopkins University

CERDEC On Energy Storage Solutions

The Future of Biometric Technology With U.S. Citizenship & Immigration Services...

Military Healthcare: Presentation Case Study

FIND CONTENT BY TYPE

IDGA COMMUNITY

ADVERTISE WITH US

JOIN THE IDGA COMMUNITY

Using Functional Populations to Create More Realistic Simulations

Upcoming Events

Follow Us

Subscribe to our Free Newsletter

Latest Webinars

Recommended

FIND CONTENT BY TYPE

IDGA COMMUNITY

ADVERTISE WITH US

JOIN THE IDGA COMMUNITY