Context situations policy Daniel Cutting Aaron Quigley University

Context situations policy Daniel Cutting, Aaron Quigley University of Sydney 19 th July 2004 Daniel Cutting

Introduction § Daniel Cutting § Ph. D. candidate at University of Sydney (Aaron Quigley supervisor, John Zic associate supervisor) § Part of the Smart Internet CRC § About half-way through Ph. D. § Thesis area: application collaboration in pervasive computing environments 19 th July 2004 Daniel Cutting 2

Outline § Pervasive computing § Motivating scenario (art gallery) § Middleware data distribution policies § Context spaces § Application to scenario § Discussion 19 th July 2004 Daniel Cutting 3

Pervasive computing § Mobile devices (constrained, wireless) + fixed infrastructure (powerful, wireline) § Hypothesis: applications in PCEs can be improved using context maximise availability of data minimise battery usage and network traffic constrained by user preferences use context to aid data distribution 19 th July 2004 Daniel Cutting 4

Art gallery scenario Bob was here. Gillian Edward Cynthia Bob was here. Sunflowers, Van Gogh

Art gallery scenario § Guide publishes data that is pushed to students (marking image of painting) § Repository shared by group stores longlived data (group photo) § Public infrastructure stores persistent data (painting images, guest book) 19 th July 2004 Daniel Cutting 6

Middleware § Publish-subscribe: good for events markings on painting image § Tuple spaces: good for data persistence guest book, group repository § Build middleware that combines the two 19 th July 2004 Daniel Cutting 7

Middleware distribution § Distributing/storing data is a problem many devices, some small, wireless may have powerful fixed infrastructure, but sometimes purely ad hoc networks § Middleware needs flexible data distribution and storage policy § Use context to aid this policy 19 th July 2004 Daniel Cutting 8

Context § Sensed/inferred values from environment, network, devices, applications and users e. g. beacons, bandwidth, storage capacity, usage patterns, preferences § Complex to base policy on raw context interpose symbolic situations context situations distribution policy 19 th July 2004 Daniel Cutting 9

Context spaces § Treat context as n-dimensional space § Each dimension is type of context e. g. [bandwidth, storage capacity] sample context vector might be [high, low] § Specific situation vectors also exist (statically specified or learnt over time) § Find “nearest” situation vector to convert context vectors to situation 19 th July 2004 Daniel Cutting 10

Context spaces Z 19 th July 2004 zz z Daniel Cutting 11

Dynamic clustering § Don’t specify situation vectors § Cluster context vectors to automatically identify inherent situations § How should policy act if no situations exist until run-time? § Situations can shift over time to reflect changes to contextual sources 19 th July 2004 Daniel Cutting 12

Scenario: context situations § Decentralised each device determines own context § To build context space, designer identifies available context, e. g. local power, bandwidth, storage neighbours’ power, bandwidth, storage size, priority, relevance, persistence of data painting beacons, etc. 19 th July 2004 Daniel Cutting 13

Scenario: context situations § Select context for dimensions data importance I, persistence P, size S context vector is of form [I, P, S] § For static space, specify situations signature, photo, demonstration e. g. photo [0. 1, 0. 8] is when data is not very important, persistent and large (like a photograph) 19 th July 2004 Daniel Cutting 14

Scenario: situations policy § A device putting data into the middleware system can: store locally, broadcast digest § Make distribution policy using situations signature broadcast photo digest demonstration store 19 th July 2004 Daniel Cutting 15

Scenario: context policy Group photo at Sunflowers Edward Gillian Group photo at Sunflowers Nearest situation vector is photo digest Cynthia Unimportant (0. 2) Long-lived (0. 7) Large size (0. 9) Bob

Discussion § Representing nominal and cyclic dimensions is troublesome § Can situations policy be automated in clustered context space? § Unknown values in context vectors could cause spurious results - project to lower dimensions? 19 th July 2004 Daniel Cutting 17

Static classification § During design-time manually specify situation vectors § During run-time measure raw context determine context vector find nearest situation vector based on a metric such as Euclidean distance space is not altered - essentially a lookup 19 th July 2004 Daniel Cutting 18