B.R. Badrinath Badri,

1

• B.R. Badrinath (Badri),
• Collaborators Tomasz Imielinski, Rich Martin,
  Brett Vickers
• http//www.cs.rutgers.edu/dataman
• badri,imielins,rmartin,bvickers_at_cs.rutgers.edu
• Funded in part by DARPA (SENSIT), CISCO

2
Scenario
Whats around me?
Where is the TA for 352?
3
Vision

• As users move through physical space, they are
  augmented with information about their
  surroundings
• Problems
• How to address, query, and gather data from a
  massive network of sensors embedded in the
  physical space
• Dataspaces
• How to organize, present, and manage rapidly
  changing information about physical space
• Infodispensers
• How to automatically construct useful indexes ---
  maps --- for data distributed in a network of
  elements (some of them mobile)
• Spatial web
• How to manage such a large scale network
• Superscale network management
• Tools for characterizing performance

4
Approach

• Build an infrastructure that will be able to
  provide an enhanced view of the surrounding
  physical space
• As users navigate physical space, they will be
  sprinkled with information (illuminated with
  information)
• Key Idea Closely tie location, communication
  (network), and information to form a spatial web
• Every data item has a scope (region over which
  it is valid)
• ltTA for 352 , Room 345gt
• ltBus departed at 352, Metlars lane busstopgt
• Maintain spatial links to nearby data
• Answer queries about physical space by searching
  or crawling the spatial web

5
Main Elements

• Main elements of Digital Sprinklers
• Dataspaces
• Scale query methods by using network primitives
  (broadcast, multicast, anycast, geocast Navas
  and Imielinski 97, gathercast)
• Infodispensers
• Collect, aggregate and distribute data based on
  spatial relevance
• Resolution inversely proportional to distance
  from epicenter
• Spatial web/landscape database
• Automatic indexing of spatial information
• Crawl physical space to infer properties

6
Digital Sprinklers Architecture
Landscape Database
Infodispensers
SuperCluster
Dataspaces
Sensor Network
7
Infodispensers

• Sprinkle/Sniff information based on spatial
  relevancy
• Disseminators/aggregators of information
  collected from dataspaces/sensors
• Users who pass by will be sprinkled with
  information
• Users can also park information on digital
  sprinklers graffiti
• Assist in answering aggregate queries
• Aggregate query on physical space ? contact
  surrounding infodispensers
• Query decomposition
• Which infodispensers to contact?
• Spatial resolver directory (where is what?)
• Location tags
• Will depend on users vector (direction, speed)

8
InfoDispensers

• Landscape populated with InfoDispensers that
  have information about the surrounding area.
• Information vending machines
• Spraying (spatially constrained) /sniffing
  Information to users who pass-by!

I have partial knowledge Need to contact others
I have complete knowledge
Beyond my knowledge Need to find out who knows
about X
Examples Who are all in the room? Is badri in
the room? What is the stress level on this bridge?
9
Infodispensers

• Local sprayers of information
• New business models
• Mom and pop Infovending machines
• Infodairy queens or Info7eleven stores

Last bus left 10 minutes ago Next bus expected in
2 minutes
10
Information dissemination
Disseminated data
Local partial
remote
Locally gathered data
11
Data/query possibilities

• Locally gathered data
• When did the last bus leave?
• Locally disseminated data
• What is the schedule for busses leaving this stop
• Local remote gathered data
• Has the last bus that left this stop reached the
  next stop
• Remote gathered data remote disseminated data
• How late are busses arriving at the next stop
• Locally disseminated data remote disseminated
  data
• What is the scheduled travel time between this
  stop and next stop
• .

12
Infodispensers

• Query optimization
• Evaluate data in a larger spatial cube, resolve
  spatial containments
• Determining query plans (order of operators) for
  a moving user
• Caching of data
• How far should data be cached?
• Use spatial relevancy (spatial distribution of
  data access)
• What to report/update?
• Not every update needs to be sent to the
  infodispenser
• Only exceptions reported (based on prediction
  models)
• Challenges
• Spatial resolvers, location tags, query
  execution, resolving proximity (5 mph vs 60 mph),
  resolving granularity, distribution of updates,
  prediction models

13
Spatial Web

• Motivation
• Query the physical space
• Inspiration
• Web is an ad-hoc structure on conceptual space
• Millions and Millions of producers
• My pages point to DCS Rutgers, Berkeley,
  Princeton, Yale, who point to
• Rich theoretic structure based on social network
  research
• Can we build a massive, ad-hoc representation of
  physical space?
• Anyone can add to the structure
• How to automatically build useful
  representations?
• Can we make meaningful queries against the
  spatial structure?

14
Physical Space as a graph
Spatial Web
15
Physical space as a graph

• Nodes or pages have embedded location tags
• Badriin ltscopeRoom 345gt
• Pages have spatial links ltsref, URL (location
  tag)gt
• Badriin ltscope ltRoom 345gt ltsref, Room346gt ltsref
  Coregt
• Tags resolve to a spatial representation
• Build spatial index by aggregating spatial
  represenatations obtained by crawling the
  surrounding physical space

16
Example Finding a house
Spatial Web
4 Sale 81 Elm St
4 Sale 2 Maple St.
17
Spatial web

• Establish a spatial link structure on surrounding
  dataspaces
• Self-organizing web of links that correspond to
  the physical space
• Physical space represented by a graph
• Answer queries about surroundings by crawling
  local space
• Link information based on spatial proximity
• Answer queries by crawling
• Crawl using these links to obtain a semantic
  structure of the physical space automatic
  construction of spatial indexes
• Trade-off accuracy for time-to-crawl
• Challenges
• Crawling while on the move, on-line crawling vs
  offline crawling, prefetching, predicting
  trajectories, transforming web structure to
  spatial web structure and vice versa

18
Motion through space
Stationary units define a stable graph
Mobile units change link structure by crawling or
reassess surroundings
19
Motion through space
Stationary units define a stable graph
Mobile units change link structure by crawling
or reassess surroundings
20
Motion through space
Stationary units define a stable graph
Mobile units change link structure by crawling or
reassess surroundings
21
Summary

• Physical space described by a collection of home
  pages
• Home pages have location tags and spatial links
• Extraordinarily dynamic content, spatially
  constrained queries
• Information architecture based on infodispensers
• Sprinkle as well as gather information
• Challenges
• Dealing with massive distribution of data
• Organizing and developing structure about the
  physical space
• Answering queries by crawling space
• Network management and maintenance (performance
  characterization)