PERVASIVE COMPUTING MIDDLEWARE BY SCHIELE, HANDTE, AND BECKER

A Presentation by Nancy Shah

Vision of Pervasive Computing“Pervasive computing envisions applications that

provide intuitive, seamless and distraction-free support for their users.”

“… the applications combine and leverage the distinct functionality of a number of devices. Many of these devices are invisibly integrated into the environment.”

“… provide high quality task support while putting only little cognitive load on users as the require only minimal manual input.”

Challenges to Application Developers

Unprecedented device heterogeneity Unreliable wireless communication Uncertainty in sensor readings Unforeseeable execution environments

From static to highly dynamic Changing user requirements Fuzzy user preferences

Development of pervasive applications in a non-trivial, time-consuming and error-prone task

Design Considerations

Organizational Model Smart Environment

Spatially limited area equipped with sensors and actuators

Stationary devices Some devices are dynamically integrated Requires a powerful coordinating computer

Smart Peers Dynamically formed collection of peers Dynamic set of devices Decentralized: No coordinating computer More flexible, but more complicated

Design Considerations (cont.) Provided Level of Abstraction

Full Transparency A completely generic solution All possible application scenarios

Configurable Automation Allows developer to forget about intrinsic

problems Abstraction automates the task with some

configuration Simplified Exposure

When configuration is no longer reasonable Restricted to automated gathering of

information

Design Considerations (cont.) Supported Tasks

Spontaneous Interaction Devices communicate with each other Detect and monitor the available set of devices

Context Management Coordinate measurements of multiple distributed

sensors Fusion of data

Application Adaptation To the overall system properties To the available capabilities To the context of their users

Spontaneous Interaction

Ubiquitous Communication and Interaction Interoperability

Requires a common understanding of the semantics of shared functionalities

Three main possibilities: Fixed Standardized Protocol Set Dynamically Negotiated Protocol Set Interaction bridges

Communication Abstractions Event Heap Remote Message Invocation

Spontaneous Interaction (cont.) Integration of Heterogeneous Devices

Wide range of devices Different hardware/software Different capabilities and resources

Two different approaches Build multiple, yet compatible systems for

different classes of devices Build modular systems with minimal yet

extensible core

Spontaneous Interaction

Dynamic Mediation Peer-based Discovery

All nodes participate in the discovery together Clients can broadcast a discovery request Service providers can broadcast its services

Mediator-based Discovery Mediators manage a service registry for all

devices Clients register services with the mediator Can coordinate entries with other mediators In absence of mediator no discovery is possible

Context Management

Definition of Context Where you are Who you are with What resources are nearby

System adapts to context according to Location of use Collection of nearby people Available network and computer infrastructure

Three Classes of Applications Context-aware presentation Automatic execution of a service Tagging of context to information for later retrieval

Context Management (cont.) Acquisition and Fusion

Accuracy Sensor readings are not a discrete value May deviate for another’s measurement of the same Sensors usually specify the mean and maximum error

Freshness Once the value is read, it begins aging The rate of change determines accuracy

Methods to improve accuracy Use multiple sensors or readings Use different technologies or perspectives

Context Management (cont.) Modeling and Distribution

Costly Sharing costs with other applications or

institutions Standardized interfaces

Retrieved information must be interpreted across a number of applications

Standardized data Applications need to be able to interpret data

Context Management (cont.) Provisioning and Access

Provides suitable abstractions for applications

Three classes of context for access Identity Location Time

Application Adaptation

Challenges Integrating coordination of devices

Different environments have different devices Set of devices changes over time Different devices can be used for same task

Coordination Based on the target environment and the user Able to adapt to many changes Needs support at the middleware layer

Not the application layer

Application Adaptation (cont.) Inter-Application Adaptation

Several non-distributed applications Applications are not aware of one another Allows for reuse of traditional applications Middleware can:

Detect changes and adapt the set of applications Provide transparent transcoding services for user data Provide services to facilitate the interaction of applications

Issues Weak form of application coordination Cannot coordinate actions between applications when

needed

Application Adaptation (cont.) Intra-Application Adaptation

Single application distributed across several devices

Requires applications to run an application-defined building block

Middleware builds a model of the application using Individual building blocks Dependencies on one another Dependencies on environment

Issues Devices must be capable/willing to install additional

code

Conclusion

Pervasive Applications Provide seamless, intuitive, and distraction-free task

support Provide a set of supportive services

Influenced by overall organization and targeted level of support Heterogeneous Devices

Alleviates low-level communication issues Handles details of gathering information from large

numbers of sensors Supports changing set of devices

Intuitive Task Support is still a challenge as new devices and scenarios continue to emerge

Any Questions?

Thanks!