Towards Ad-hoc Situation Determination Graham Thomson, Paddy Nixon and Sotirios Terzis

Pervasive and Global Computing Introduction  PlaceLab has been successful in making location information freely available for use in experimental ubiquitous computing applications.  We envisage a need for tools that can deliver a much richer set of contextual information.  The high-level situation of the current environment is a key contextual element.

Pervasive and Global Computing Scenario  Jane captures a ‘coffee break’ snapshot on her smartphone and instructs it that both calls and messages should be announced audibly when she is in that situation.  Historical snapshot of ‘formal meeting’ captured from context server.  In a partner company’s building her smartphone is unable to determine the situation. The local context server reveals it is ‘formal meeting’, and the previously associated behaviours are applied.

Pervasive and Global Computing Drawbacks of the state of the art  Require an environment expert.  Reasoning mechanisms must be manually constructed and maintained.  Difficult to specify sensor to situation correlations on a large scale.  Situation specifications will suffer from the subjective bias of the expert.  Reasoning is performed by a single oracle which cannot exploit non-public knowledge.  Once programmed, situations are fixed.

Pervasive and Global Computing Drawbacks of the state of the art  Require an environment expert.  Reasoning mechanisms must be manually constructed and maintained.  Difficult to specify sensor to situation correlations on a large scale.  Situation specifications will suffer from the subjective bias of the expert.  Reasoning is performed by a single oracle which cannot exploit non-public knowledge.  Once programmed, situations are fixed.

Pervasive and Global Computing Representing the situation  Contextual information is captured as a relation between two instances of a class.  Ubiquitous computing environments are open - any number and variety of people, devices, and software may appear within them.  Contextual information they produce are therefore also open, as the instances of a relation are drawn from a potentially infinite set.  Reasoning about contextual information is then made difficult, as many machine learning techniques make strict demands on the structure and constraints of the data, e.g. the C4.5 algorithm.

Pervasive and Global Computing Representing the situation  Text classification aims to automatically assign documents to a given set of categories.  Documents may exhibit no regular structure.  The set of known terms used in the documents may grow with each new document that is classified.  Easy to see similarities between the task of text classification and that of situation determination.

Pervasive and Global Computing Situation as a document  Document  Bag of terms.  Situation  Bag of relations.  A relation is expanded to facilitate reasoning at an abstract level. JaneRobert works with

Pervasive and Global Computing Situation as a document Person JaneRobert works with Robert works with Jane works with Person

Pervasive and Global Computing Situation as a document  To reason about snapshots, we must transform them into a representation suitable for machine learning algorithms.  Use vector space model - considers a document to be a vector in a multi-dimensional Euclidean space, with each axis corresponding to a term.  Extra time axis added to vector, scaled from 0 (midnight), to 1 (just before midnight).  Approach based on Support Vector Machines (SVM), which are currently the most accurate classifiers for text.

Pervasive and Global Computing Determining the situation

Pervasive and Global Computing Determining the situation

Pervasive and Global Computing Determining the situation

Pervasive and Global Computing Determining the situation

Pervasive and Global Computing Determining the situation

Pervasive and Global Computing Drawbacks of the state of the art  Require an environment expert.  Reasoning mechanisms must be manually constructed and maintained.  Difficult to specify sensor to situation correlations on a large scale.  Situation specifications will suffer from the subjective bias of the expert.  Reasoning is performed by a single oracle which cannot exploit non-public knowledge.  Once programmed, situations are fixed.

Pervasive and Global Computing Process of interaction  Determining the situation within an environment is a cooperative effort between each participant (device) within it.  Each participant’s view is the union of its private, privileged, and public relation sets.  Make simplifying assumption that participants are in the same situation if they are in the same room.  Interaction proceeds as in the following five steps:

Pervasive and Global Computing Process of interaction Stage 1 – Proposal of current situation Formal meeting Formal meeting Formal meeting Formal meeting Coffee break

Pervasive and Global Computing Process of interaction Stage 2 – Sharing of relevant snapshots

Pervasive and Global Computing Process of interaction Stage 3 – Correction (freshness, confidence, accuracy, derivation algorithm) P = 2.3 P = 2.35 P = 1.8 P = -34 P = ?

Pervasive and Global Computing Process of interaction Stage 4 – Heuristic selection (strongest non- obvious indicator) Formal meeting

Pervasive and Global Computing Process of interaction Stage 5 – Learning (dynamic adaptation to drift) Formal meeting

Pervasive and Global Computing Drawbacks of the state of the art  Require an environment expert.  Reasoning mechanisms must be manually constructed and maintained.  Difficult to specify sensor to situation correlations on a large scale.  Situation specifications will suffer from the subjective bias of the expert.  Reasoning is performed by a single oracle which cannot exploit non-public knowledge.  Once programmed, situations are fixed.

Pervasive and Global Computing Preliminary results  Situations:  Formal group meeting  European project meeting  Informal meeting  Coffee break  Private study  Movie night  Ubicomp environments are highly dynamic - new situations will continually appear, while current situations will cease to recur.  The system must achieve an acceptable determination accuracy using as few snapshots as possible.

Pervasive and Global Computing Preliminary results

Pervasive and Global Computing Preliminary results  Almost all misclassifications occurred between the three meeting situations.  THANK YOU!