How to Build Smart Appliances: ContextAware System

1
How to Build Smart Appliances Context-Aware
System

• January 14, 2004
• Choi, Chang Ik

2
Overview

• Introduction
• Terminology
• Context Capturing Devices
• Constraints on Capture
• Architecture
• Building a Context-Aware Application
• Example
• Research Areas
• Thoughts

3
Introduction

• Smart appliances should
• Not be ignorant about the environment
• Have implicit human-computer interaction
• Work naturally

4
Terminology(1/2)

• Context Awareness
• knowledge about the users and devices state,
  including surroundings, situation, and location
• Situation (situational context) real-world
  situation
• Sensor data (situational data) data that is
  captured to represent the situation
• Context (context knowledge) abstract
  description of the real-world situation
• Context-aware/sensitive application
  applications that change their behavior according
  to the context (independent of sensor data)

5
Terminology(2/2)
Abstract representation of the situation
Real-world situation
Application
Abstraction
Sensor data
Context
6
Context Capturing Devices

• Light Sensors
• supply information on the light density,
  intensity, reflection, etc. of the device
• Accelerometers
• supply information on the inclination, motion,
  or acceleration of the device
• Location
• GPS, WLAN, GSM, etc.
• Passive IR Sensors (Motion Detector)
• for stationary devices movement of the device
  itself is detected as well

7
Basic Assumption of Sensor-Based
Context-Awareness

• In a certain situation, specific sensor data is
  frequently similar to sensor data of the same
  situation at different times

8
Constraints on Capture

• Portability, Usability, and Design
• Power consumption
• Calibration
• Setup Time
• Reliability
• Price and Cost
• Unobtrusiveness

9
Building a Context-Aware Application

• Identify the contexts that matter check if
  context matters at all (select a situation)
• Find that appropriate sensors (select a context)
• Build and assess a prototypical sensing device
• Determine recognition and abstraction
  technologies (independent of the real situation)
• Integration of cue processing and the context
  abstraction
• Build applications

10
Architecture (1/3)
(Buffer)
(Buffer)
11
Architecture (2/3)

• Cues
• Provide an abstraction of physical and logical
  sensors
• Make changes of the hardware transparent to the
  context recognition layer
• Reduce the stream of data provided by the sensors
• To implement cues, use statistical functions to
  get summary of the values over time or to extract
  features form the raw data
• Average, standard derivation, quartile distance,
  base frequency, first derivative, time domain
  values

12
Architecture (3/3)

• Contexts
• Description of the current situation on an
  abstract level and is derived from available cues
• e.g.) If (light is low and acceleration is
  nonzero) then context Moving in the dark
• AI methods and mechanisms are needed for
  calculating context from the cues

13
Example (A Context-Aware Mobile Phone)

• Multiple configurations (profiles)
• Hand vibrate mode
• Table gentle sound
• Pocket loud sound
• Outside sound, as high as possible
• General when none of the above apply

14
Research Areas

• Cue Tuple Space to Context
• Finding the right cue for the context
• Context Tuple Space to Application and Scripting
• Finding the right context or group of contexts
  for an application

15
Thoughts

• Considerations for selecting sensors and
  algorithms
• Power Consumption
• Size and Weight
• Price of Components
• Robustness and Reliability
• Discriminating contexts that should be dealt
  implicitly and contexts that should be dealt
  explicitly