Hovering Information
Introduction
Hovering information is a mobile
computing paradigm where pieces of self-organising information are
responsible to find their own storage on top of a dynamic set of mobile
devices. Once deployed, the hovering information service is a
location-based service disseminating geo-localised information
generated by and aimed at mobile users. It supports a wide range of
pervasive applications, from urban security to stigmergy-based systems.
A piece of hovering information is attached to a geographical point,
called the anchor location, and to its vicinity area, called the anchor
area. It is responsible for keeping itself
alive, available and accessible to other devices within its anchor
area. It does not rely on any central server. Simulation results so far
show that our replication and caching algorithms scale up to 200
distinct pieces
of hovering information concentrated in a small geographic area.
Examples
1. Virtual tags are inserted at
specific locations on roads or
motorways either by cars' drivers or traffic management staff. The
purpose being to provide information to cars' drivers about road
conditions, accidents, etc. In such a scenario, using the notion of
hovering information, such tags will not be stored on a specific server
and made available to users when they reach the zone of interest of the
information. Instead the tags are locally stored in the cars and made
available through wireless channels
to nearby cars. Since data have a meaning for the specific
location
they have been attributed, data will have to "change" car as soon
as the car they are
currently stored in leaves the area of the anchor location. The data
will then hop from one car to the next one.
2. In an emergency scenario, virtual
data present before a disaster
may want to "survive" by using emergency crew or survivors
devices.
This data can also present useful information for emergency services.
Additionally, disaster's
survivors may want to indicate their position by placing the
appropriate hovering information attaching it to their own location.
Emergency crew member can place hovering information to areas where
survivors have been found or where there is a chance to find some
survivors. In this case, the information will hop from one
emergency/survivor device to another one.
Current
Work
- Hovering
Information model: we have defined a model of user-generated
data that stores itself on mobile devices keeping itself available at a
specific location, including dependability requirements.
- Replication
and Caching algorithms: pieces of hovering information follow an
Attractor Point Algorithm and get attracted to their anchor location.
Nodes implement the Location Based Caching algorithm rank data based on
their distance to their own anchor location and discards the farthest
data.
- Simulations: we have performed a series of simulations
using the Omnet++ environment involving up 200 pieces of hovering
information and up to 200 nodes. Mobility patterns follow a RandomWay
Point mobility model. Results show that the Attractor Point Algorithm
combined with the Location Based Caching algorithm is scalable in terms
of the number of pieces of hovering information that the system can
support. They also show the emergence of a loadbalancing property of
the buffer usage which stores replicas in an equilibrated and optimal
way as the number of pieces of hovering information increases.
- Real
mobility patterns: We have tested the algorithms under a
RandomWay Point mobility model and under ideal wireless conditions.
This is not characteristic of real world behaviour. We are currently
applying the different algorithms to scenarios following real mobility
patterns (e.g. crowd mobility patterns in a shopping mall or traffic
mobility patterns in a city) with real wireless conditions (e.g.
channel interferences or physical obstacles).
- Spatial
Memory: We are currently defining and implementing a distributed
memory service for storing and retrieving pieces of hovering
information to and from geographic regions, exploiting available
(stationary and mobile) devices as the main storage medium.
Papers
- A. A. Villalba Castro, G. Di Marzo Serugendo, D.
Konstantas. Hovering Information -
Self-Organising Information that Finds its own Storage. Autonomic
Communication Vasilakos, A.;
Parashar, M.; Karnouskos, S.; Pedrycz, W. (Eds.) Springer,
2009. (pdf)
- A. A. Villalba Castro, G. Di Marzo Serugendo, D.
Konstantas. Hovering Information. Activity-Based Computing (Works In
Progress), IEEE Pervasive Computing 7(2):58-61 (2008) (pdf)
- A. A. Villalba Castro, G. Di Marzo Serugendo, D.
Konstantas. Hovering Information -
Self-Organising Information that Finds its own Storage.
International IEEE Conference on Sensor Networks, Ubiquitous and
Trustworthy Computing, Taichung, Taiwan, 11-13 June 2008, pp. 193-200. (pdf)
- A. A. Villalba Castro, G. Di Marzo Serugendo, D.
Konstantas. Hovering Information :
Infrastructure-Free
Self-Organising User-Generated Location-Aware
Information Dissemination Service.
Second ERCIM Workhop on eMobility, Tampere, Finland, 2008. (pdf)
- A. Villalba Castro, G. Di Marzo Serugendo, D. Konstantas, "Hovering Information -
Self-Organising Information that Finds its Own Storage",
BBKCS-07-07, Technical Report, School of
Computer Science and Information Systems, Birkbeck College, London, UK,
Nov 2007.
(pdf)
- G. Di Marzo Serugendo, A. Villalba Castro, D. Konstantas: "Dependable
Requirements for Hovering Information", Supplementary
Proceedings of the 37th Annual IEEE/IFIP International
Conference on Dependable Systems and Networks (DSN'07), June 2007. (pdf)
Poster
- Hovering Information. 50 years of Computing at Birkbeck. (pdf)
Partners
- University of Geneva
- Birkbeck College
G. Di Marzo Serugendo
Sep 2009