Overview
This proposal therefore addresses three key research areas related to the understanding and management of massive human crowds:
- Data Collection (Wireless bracelets, UAVs, cell phone, CCTV etc.)
- Data Analysis (Real-time)
- Modelling and Prediction
data collection
|
Whether it is planning for large events, or analysing egress patterns, one of the biggest challenges in understanding human crowds is access to large volumes of high quality data. Capturing individual dynamics at the spatial and temporal resolution necessary for real understanding is very challenging. For the experiment we plan to create purpose built devices (bracelets) that are capable of detecting and recording interactions with other nearby devices. From this data it is possible to extract longitudinal proximity graphs, defining exactly which people are close together and for how long they are close together. Beyond the wristbands we will also use additional data, in particular video from drones, cell phone data, human observation and more. From the collective data sources we hope to distill brand new insights regarding human crowd dynamics.
|
|
DATA ANALYSIS
The planned data collection process will involve three disparate data sources, wristbands, cell phone (CDRS analysis) and video recordings. Each of these raw data sources will require completely different methods of processing, and the resulting analysis will likely be at different spatial and temporal scales. Some of the data sources, because of the nature of the collection process or the complexity of the analysis, will lend themselves more easily to the possibility of real-time processing (or at least very low latency). Finally, the idea will be to fuse the data sources with three goals in minds. Firstly, the data feeds can be used to examine patterns in the crowd and look for early warning signals of pending danger. Secondly, this data will provide information on human crowds at unprecedented scale and resolution. This can lead to fundamental new understanding of the dynamics of human crowds, for example identifying the densities at which the impact of individual behaviour on macro level dynamics start to diminish (does this happen as a form of phase transition). Finally, the data should be used as input to the simulations.
Modelling and simulation
|
Human crowds are classic examples of complex systems, where the small local interactions between people can lead to emergent macro-scale dynamics that can be extremely hard to predict. For this reason modelling and simulation has been of the most popular approaches for understanding, predicting and managing crowds. In this proposal we plan to develop new models and simulations that are in principle capable of dealing with: rapid parameter space exploration (for scenario planning) and crowds of more than 100k people. These challenges will be addressed through high-performance distributed computing.
|
|
