Paul M. Torrens

Why would you model crowds?

Important factor of modern life. Some of the most important tipping points. Evacuation, emergencies. Understanding how crowds work is important for public health. New forms of mobs.

“We [really] do not know as much about crowds as we would like to know.”

“Simultation can serve as an artificial laboratory for experimentation in silico

“I build complex systems of behaviourally-founded agents, endowed….

“What does this have to do with geospatial technology?”

Business opportunities for Geographic Information Technologies”

Way to do it is convoluted, build models, AI, rendering, outputting analysis, statistics, GIS, etc.

Have technology to model individual people, often don’t have the data. Generate synthetic populations, downscaling larger data sources to individual level. Gives characteristics to our agents. Customer loyalty cards, GPS, cell phones.

Physical modelling and rendering

“Small-scale geography from motion capture and motion editing”

Record spatial and temporal information in studios, 100 frame per second, spatially in order of a few cm. Graph of movement of skeleton through time.

Physical simulation, doing bad things to the modelled skeletons.

“Behaviour is simulated (!= scripted) using computable brains”

Using a turing machine, socio-communicative emotional agent based model, wrapped in geographic information. Given GIS functionality. Geosimulation.

Taking basic model and wrapping in geography. Agents can “see”, can deploy mental map, plan past, parse to waypoints, navigate to goals. Identify what they’re interested in, they can steer, locomote. Use motion capture date.

Should be able to drop them in a city and they’ll get going.

Data relies on GIS, video of a 3D model of a city. Space-time signatures, space-time patterns. Given someone’s usual geo behaviour can figure out all their possible directions.

Social network monitoring. Monitored children on a campus every day for three years, watched how they formed groups and how they play. High performance computing. Binary space partitions.

“Applying this to real world issues”

“Quotidian crowd dynamics”

Screen scenes. Pick one person and follow them.

Showing video, model of people. Old people, young people, drunk people. Each behaving autonomously. Positive feedback, negative feedback, all sorts.

“Extraordinary scenarios”

Building evacuations, bottlenecks.

Run through space-time GIS, look at egress behaviour, if they run more people get hurt.

Urban panic, out of buildings into urban environment, look at how they evacuate. Showing video people jogging.

More diagrams and graphs.

Dynamic density map.

Riotous crowds. Standard riot model and wrapping with geo-spatial exo-skeletons. Generates a riot. Can test for clustering to see if people with similar motives and emotions are grouping. Small scale riot behaviour turns to large scale very easily. Devious behaviour, rioter see police and pretend they’re not rioting, run away when chased. Inserting police who are told not to arrest people will calm the crowd but not completely.

Crowd response to invasion of non-native stuff. Showing Cloverfield.

Small scale epidemiology of influenza.

Zombies.

Putting crowds into digital environment, what happens to location based services.

geosimulation.org

Modeling Crowd Behavior

Technorati tags: geosimulation, simulatation, modelling, crowds, where, where2.0, where2008