978-1-4244-3304-9/09/$25.00 ©2009 IEEE
Using Situation Lattices in Sensor Analysis∗
Juan Ye, Lorcan Coyle, Simon Dobson, and Paddy Nixon
System Research Group, School of Computer Science and Informatics,
UCD, Dublin, Ireland
E-mail: juan.ye@ucd.ie
Abstract
Highly sensorised systems present two parallel chal-
lenges: how to design a sensor suite that can efficiently and
cost-effectively support the needs of given services; and to
extract the semantically relevant interpretations, or “situa-
tions”, from the flood of context data collected by the sen-
sors. We describe mathematical structures called situation
lattices that can be used to address these two problems si-
multaneously, allowing designers to both design and refine
situation identification whilst offering insights into the de-
sign of sensor suites. We validate the accuracy and effi-
ciency of our technique against a third-party data set and
demonstrate how it can be used to evaluate sensor suite de-
signs.
1. Introduction
A pervasive computing environment assumes a number
of invisible sensing/computational entities that collect in-
formation about users and an environment. With the help of
these entities, a pervasive computing system will be able
to understand users’ situations or demands so that it can
deliver services in a contextual manner. To facilitate this
understanding, we propose a well-founded mathematical
structure, called the situation lattices, to study relationships
between a large number of low-level sensed data and high-
level situations or human activities.
Within a situation lattice, sensor data are abstracted and
organised with respect to their semantics (such as different
abstraction levels and conflicts between them). The situ-
ation lattice will apply basic semantics to learn richer se-
mantics that describe relationships between sensor data, be-
tween sensor data and situations, and between situations.
∗This work is partially supported by Science Foundation Ireland under
grant numbers 05/RFP/CMS0062 “Towards a semantics of pervasive com-
puting”, 07/CE/I1147 “Clarity, the centre for sensor web technologies”,
and 04/RPI/1544 “Secure and predictable pervasive computing”.
Situation lattices have been proposed in our earlier
work [22, 23], which allowed developers to manually or-
ganise abstracted sensor data in a lattice and to label them
with situations. The preliminary evaluation results were
promising when we constructed a simple situation lattice
to describe research activities in an office environment by
using a small number of sensor data. This paper extends
the initial model of situation lattices to deal with more com-
plicated environments where a larger number of sensors are
involved and relationships between sensor data and situa-
tions are less explicit.
This paper proposes new approaches to automatically
generate a situation lattice with respect to context predi-
cates, which are abstracted from sensor data; and to auto-
matically label context predicates with situations. We will
use situation lattices to help system developers to evaluate
the performance of sensors, including evaluating whether
a new sensor (or a new type of sensors) should be intro-
duced, depending on whether existing sensors can suffi-
ciently recognise situations.
To test the general applicability of a situation lattice, we
demonstrate its feasibility by building it from the third-party
real world data set – the PlaceLab data set [7]. The Place-
Lab is an instrumented home that contains over nine hun-
dred sensors. The data set was gathered in the real world
conditions in that a married couple (who were unaffiliated
with the PlaceLab research) lived in the PlaceLab over a
period of 15 days. During this period, they were encour-
aged to maintain their life routine as normal as possible.
The Placelab was also instrumented with the audio-visual
recording infrastructure that was used to record activities of
the subjects except for private activities (such as bathing).
The video was annotated by a third party, which provided
the ground truth of this data set. So far, only the activities
of the male subject were annotated. Besides the sensor data
and annotated diary, this data set involved a location map
of this home, and a sensor metadata file. This metadata
file records the meta information about each sensor input,
including its type, its identity (ID), where (e.g., the living