This Extreme Wireless Distributed Systems (EWiDS) project takes its inspiration from managing and controlling large crowds of people. Its uniqueness lies in the way observations take place. Traditionally, crowds are monitored through a few tens of cameras placed at strategic locations, leading to global, external observations. In this project we assume that thousands of sensing devices are present inside a crowd, leading to local, internal observations that are not geographically restricted to a specific instrumented area. Such a sensing device is typically a smartphone (now owned by many people) or a low-cost electronic badge. By harnessing devices people already own we would be able to monitor crowds without large infrastructure investments. Our approach however needs to overcome extreme networking conditions that have not been tackled before: large scalability, extreme high densities, mobility and real-time performance.
The scientific ICT challenge for EWiDS is to automatically extracting and analyzing the evolving structure of large groups of (autonomous) mobile entities, be they people as in crowds, tagged cargo as in logistics or bicycles in a city. By detecting which of sensor devices are close to each other, and when, we obtain a picture of the group from within the group; a picture that continuously changes as sensor nodes move: a proximity graph. An important specific ICT challenge is transferring massively distributed continuous streams of low-level, highly faulty radio-based proximity detections from very simple sensor nodes to an external analysis system, transforming those streams into a highly reliable and continuously up-to-date view on dynamic proximity graphs.
Within the proximity graphs, scientific questions arise: Can we (automatically) discover lanes, clogging, or other structural properties? Is it possible to discover conversing groups or other socially related structures? And, given feedback from users, can we say anything about the mood in (parts of) a crowd?
Biggest results so far
Wireless crowd monitoring in Arnhem
Monitoring the movements of crowds in cities can lead to improved city planning, more efficient traffic flows and safer crowd management. As camera surveillance might lead to privacy violations, we use wireless sensor networks to measure who is close to whom. In particular, we use ordinary smartphones as sensors. We make sure that the data of individuals are anonymized. More.
ICT science question: how can we reliably detect mobile devices and realistically project their trajectories onto a city plan? One of the problems is that there are many false and missed detections, originating from very different sources. Identifying trajectories is difficult as there may be many alternatives paths between two subsequent detections of the same device at different locations.
Involved COMMIT/partners: Gemeente Arnhem, Wireles Arnhem, VU.
Measuring crowd densities for safety and efficiency
View presentation by clicking on this image
Events like concerts, festivals and sporting competitions often attract a crowd of people. The same can be the case for institutions like museums, hospitals and amusement parks. We have developed a real-time visualization of how the density of a crowd changes. More.
ICT science question: the scientific challenge is how to reliably estimate the number of people that are in the neighbourhood of each person. Each person is a node in a constantly changing network. This estimation is a scientifically hard problem, because we consider mobile networks with high densities: each node has typically hundreds of neighbours.
Involved COMMIT/partners: Van Mierlo, VU, TUDelft.