Personalized Navigator for the University of Southern California Dheeraj Kota, Neha Laumas, Saurabh Sonalkar, Urmila Shinde, Karthik Dantu, Sameera Poduri and Gaurav Sukhatme

1
Personalized Navigator for the University of
Southern California Dheeraj Kota, Neha
Laumas, Saurabh Sonalkar, Urmila Shinde,Karthik
Dantu, Sameera Poduri and Gaurav Sukhatme
USC map
2
Agenda

• Motivation
• Introduction to deSCribe
• Related work
• Features of deSCribe
• Design
• deSCribe in action
• Limitations and Assumptions

3
Motivation

• Problem
• Hundreds of visitors unfamiliar with the
  University of Southern California campus every
  year
• Campus is very large and paper maps are difficult
  to read
• Solution
• A virtual tour guide providing
• Turn-by-turn navigational instructions from the
  current location to a requested destination
• Contextual information regarding the surroundings

4
What is deSCribe?

• Contextual Information System
• Used as a pointing device to recognize buildings
  on campus with the help of the compass on the
  phone
• Distinguishes between the floors of a building
  and provides information regarding the floor
  being pointed to.
• Both features include audio descriptions using
  Text To Speech
• A Navigation System
• Turn by turn navigation to any location on
  campus.
• Displays current location on the USC map and the
  next location the user would need to move to
  reach the destination.

5
Related Work

• Schmidt-Belz et al. report their findings of a
  user validation study for the CRUMPET system
  which is location aware mobile based tourism
  system. People need more textual/audio tour
  description as they find it difficult to
  interpret maps. Location based services would be
  very important for mobile based tourism
• Kray et al. review several navigation assisting
  services/devices Observation is that more
  situational awareness is crucial for future
  mobile guides.
• Wikitude is an augmented reality application
  which provides contextual information by using
  the HTC G-1 camera view to display annotated
  landscape, mountain names, landmark descriptions,
  and interesting stories.

6
Phone as a remote control

• Allows a user to determine the name of a campus
  building just by pointing to it with the phone
• Design
• The phone uses the compass and GPS sensor that
  are inbuilt to determine the building being
  pointed to
• Capabilities
• Gives the user the option to choose the
  granularity of details (fine/coarse) provided
  regarding the building
• Requires
• Database of all USC campus buildings including
• Building names
• GPS co-ordinates
• Description of buildings

7
Compass Algorithm
y
Distance (d1)
135 deg
x
Distance (d2)
Distance (d4)
Distance (d3)
8
Route Determination

• The user can navigate through the USC campus to
  any building on the USC campus.
• Design
• Manually created a database that includes GPS
  co-ordinates of
• The end-point of every street
• Street intersections
• Calculate the point of intersection of a
  perpendicular line passing through the current
  point and the nearest road to find nearest point
  on the road.
• A algorithm is used to provide further
  directions.
• Usability
• Directions are provided relative to the direction
  the phone is pointing
• Dynamically updates directions when user goes
  along a wrong path.

9
Floor Information

• Extracts the floor number being pointed to by the
  phone. The floor number along with a brief
  description of the floor are spoken in clear
  audio. Speaks weather information on pointing to
  the sky.
• Uses the Y axis of the compass

Height tan (Y-axis) distance Floor No ceil
(Height/Height of a floor) Height of a floor
3.7 m Get weather information using the Google
weather API
10
Map functionality

• Displays current user location (red marker) based
  on GPS co-ordinate acquired
• Indicates next hop location (blue marker) on
  route to final destination
• USC map is overlaid on Google map to allow user
  to see relevant campus information with ease

11
Results

• We collected data from 30 buildings and our
  accuracy rate was 83.2 .
• Buildings incorrectly identified were large in
  size. E.g. Parking Structures.

12
DeSCribe in action
13
Assumptions and Limitations

• Database currently identifies each building on
  campus by a single GPS coordinate. Estimation of
  the building being pointed at would be extremely
  precise if GPS coordinates identifying all four
  corners of the building were available
• GPS coordinates are often not located in the
  center of buildings and building sizes are
  variable on campus. This sometimes leads to some
  inaccuracy in the estimation of the angle the
  phone is making with the buildings in the
  vicinity and leads to incorrect estimates by the
  phone.
• Assumed that all roads on campus are linear not
  always true. This sometimes leads to slightly
  ambiguous directions being given to the user when
  roads are curved.
• GPS measurements are used extensively by our
  application. These measurements use a lot of
  power negatively impacts phone battery
  lifetime.

14
Conclusion

• Built a mobile tour guide for the University of
  Southern California
• Framework extensible to other large geographic
  areas like theme parks, campuses etc.
• Introduced the use of the phone as a pointing
  device to personalize information delivery

14/15
15

• Thank You!
• Questions?