Context-Aware Applications

Introduction

Context-aware is ware applications can discover and take advantage of contextual information such as User location, time of day, nearby people

and devices and user activity. There are now quite a few different

examples of context-aware applications but are not widely available.

We will briefly discuss several applications and issues related to defining and maintaining context.

Defining Context

Dictionary definition: “the interrelated conditions in which something exists or occurs”

One definition [Schilit]: Computing context: connectivity, communication

cost, bandwidth, nearby resources (printers, displays, PCs)…

User context: user profile, location, nearby people, social situation, activity …

Physical context: temperature, lighting, noise, traffic conditions …

In addition there is time context (time of day, week, month, year…) Context history can also be useful

Defining Context

Another view of context: “Context is the set of environmental states and settings that either determines an application’s behavior or in which an application event occurs and is interesting to the user” Active context: Influences the behavior of the

application• Location in a call forwarding application

Passive context: Context that is relevant but not critical• Active map application: display location name and other

people in the room

Defining Context

There are two ways to use context: Automatically adapt the behavior of the

application based on the context (active context).

Present the context to the user on the fly or store the context for the user to retrieve later (passive context).

Defining Context-Aware Computing

Follows from the definition of context

1. Active context awareness: An application automatically adapts to discovered context by changing the application’s behavior.

2. Passive context awareness: An application presents the new or updated context to an interested user or makes the context persistent for the user to retrieve later.

Example Application: Call Forwarding

Based on a predecessor of Active Bats System called Active Badges

Location of employees is shown to receptionist Receptionist forwards calls to the nearest phone Eventually, receptionist was removed from the loop;

calls forwarded automatically Is this a good idea?

What if you were in an important meeting? (or taking a nap!)

What if a private call is forwarded to a non private room?

What is the context? The context is the user location which is active. There is no passive context.

Example Application: Teleporting

Teleporting is sometimes called “follow-me” computing

Uses Bats Dynamically maps the user interface

onto the resources of the surrounding computing and communication facilities.

Context User’s location (active) No passive context

Example Application: Active Map

For Xerox Parc Location information is collected directly

from the “Parc Tab system” Individual faces are shown at the

locations of people on the map which is updated every few seconds, allowing people to be located quickly or to notice a meeting one might want to attend.

Context: User’s location (passive) No active context

Example: Shopping Assistant

Scenario: You are at the Home Depot trying to figure out what you need to finish your basement

Shopping assistant can Tell you what parts you need Where to find them relative to your location in the store What is on sale Do comparative pricing Use your previous profile information to customize

shopping and delivery

Context Customer’s location within the store (active) No passive context

Example: Augmented Reality

User’s view of the real world is augmented with additional information

Scenario 1: You wear sun-glasses with a display and headphones and walk in downtown Rome As you move around, the glasses can tell you your

location As you look at the coliseum you get information about it

Scenario: Special ops squad infiltrating an enemy complex wearing the same gizmos In addition to information as above, the glasses change

to night vision if lighting is not sufficient

Context Location including orientation (active)

Example: Adaptive GSM phone/PDA

PDA: notepad application changes its characteristics depending on user activity Large font when walking, small font when stationary Change the intensity level depending on the lighting

conditions

Phone: decide ring volume or vibration depending on situation In hand, in a suitcase, on a table, in a

classroom/conference?

Context: User’s activity, light level pressure and proximity of

other people.

Example Application: Office Assistant

The assistant is an agent that interacts with visitors at the office door and manages the office owner’s schedule.

Assistant is activated when visitor approaches (detected by two pressure sensitive mats placed on both sides of the office door).

Assistant adapts behavior based on visitor identity, office owner’s schedule and busy status and willingness of owner to see visitor.

Context Office owner’s current activity and schedule (active).

Example Application: Where am I?

(Active map)

What’s near me? Find this for me

(Resource discovery)

“Print map on a color printer,” and system sends data to nearest available free color printer and tellsyou how to get there Location by “intent”

How do I get to Jorg’s office?

Observations

Location is the most heavily used context A natural consequence of these being mobile

applications?

Why aren’t other contexts used more heavily? Not that useful? Difficult to sense? Need more imagination?

No killer app.; little gizmos of limited use Still to come?

Sensing Low-Level Contexts Beyond Location

Time Easy to collect from the built-in clock of the

computer Nearby objects

If the system records the locations of people and other objects, it is relatively easy to figure out who and what are near us by just querying the location database.

Sensing Low-Level Contexts Beyond Location

Specially designed sensors can sense some low-level types of physical context. Examples include: Photodiode to detect light level IR sensor to detect the proximity of humans A microphone to detect sound Possible to build sensors for temperature

Sensing High-Level Contexts

Raw contextual information includes location, noise level and temperature.

An example of high-level context information is user’s “current activity” Difficult to sense complex social contexts – Example:

Lots of people in a room could indicate a party or a wake.

One approach is based on machine vision which is uses camera technology and image processing.

Another approach is to consult the user’s calendar directly to find out what the user is suppose to be doing.

Yet another approach is to use AI techniques to infer context by combining several simple low-level sensors.

Sensing Context changes Publish subscribe model?

Monitor polls the current context Notifies subscribers to the context Centralized or on a per-node?

Polling rate is a function of rate of change of context

Modeling Context Information How do we answer questions such as:

Given an object what is its location Given a location, return the set of objects

there Determining paths between locations

Security and Privacy IssuesDo we really want our location to be

known all the time?

Symbolizing Location

Applications may require a wide variety of location representations. Precise coordinate systems will be needed

for some applications. There may be higher levels of abstraction

needed depending on the application semantics.

Modeling the Environment

Detailed model describing entities in the real world and their possible interactions.

This model sets out the types, names, capabilities and properties of all entities which are relevant to context-aware application.

The extent of the data model defines the limits of the context-aware system’s view of the world and thus the possible applications.

We will now look at one approach described in “The Anatomy of a Context-Aware Application”.

Systems Infrastructure

Use CORBA and databases to implement persistent distributed objects

Objects are stored in the database as rows of data and associated operations are in SQL.

Accessed by a proxy server that receives CORBA calls.

Some information is updated too frequently to be stored in a database (e.g., information from Bat sensors)

Possible for thousands objects; proxies may want to cache.

Updating the Model

To populate the database, information about real world objects and their properties must be gathered.

Elements that frequently change, need to have those changes detected and entered into the database automatically (or as much as possible).

Updating the Model

Example: System information such as keyboard activity or machine load.

Three classes of resource monitors: Machine activity e.g. keyboard activity Machine resource e.g. CPU usage, memory

usage Network point-to-point bandwidth and latency

Used to provide a windowing-system-independent means for a user to access his desktop environment from any networked machine.

Client level event filters

Thought must be put into the design of resource monitors to ensure that the database does not become a bottleneck

Update Frequency The frequency at which items are monitored is based

on how quickly the item tends to change

Relevancy If a value has not changed significantly, it is not sent.

This value depends on the data being monitored

Caching Caching improves performance at the cost of

consistency

Location-Aware API

Absolute and relative spatial facts “Person is at (x,y,z) facing in direction ” .vs. “Person

is standing in front of the monitor” Ideally, it is not the task of the application to

translate absolute spatial facts into relative spatial facts.

To automate the process of translation, a simple method of formalizing relative spatial facts is required.

Approach: Express relative spatial facts about objects in terms

of geometric containment relationships between spaces associated with those objects.

Location-Aware API

Geometric containment is used for relative spatial facts

monitor person

Contained(person, screenspace)

Location-Aware API

Scalability With many devices, containment is complex Use a containment tree indexing system (a

quad-tree based approach)

Location-Aware API

Interaction between applications and spatial monitors Applications associate a space with an object type Applications register callbacks for given spatial facts

(e.g., when a person space becomes contained by a workstation space)

Location events are generated Location events are associated with objects. The spaces associated with a given object is looked

up which generates a stream of space location events.

This is used as input to an indexing system which generates stream of containment events.