Context Reasoning and Prediction in Smart Environments: the Home Manager case
1. Context Reasoning and Prediction
in Smart Environments:
the Home Manager case
IIMSS 2017
Algarve - Portugal, 23 June 2017
Roberta Calegari, Enrico Denti
2. Outline
ā¢ Scope & Goals
ā¢ Smart Environments in the Butlers perspective
o The Butlers Vision
o The Home Manager Platform
o The Home Manager Architecture
ā¢ Smart Environment in HM
ā¢ The Smart Kitchen Case Study
o The Smart Kitchen: Scenario
o The Smart Kitchen: Prototype
ā¢ Conclusions & Future Work
3. Scope & Goals (1/2)
Smart Environments socio-technical nature calls for
from diverse fields ļ multi-paradigm perspective
REQUIREMENTS
ā¢ availability of an effective coordination middleware
ā¢ effective support to situatedness
ā¢ guidelines and enabling techniques exploiting concepts,
methodologies, technologies from the most diverse fields,
in a multi-paradigm perspective
ā¢ skills
ā¢ concepts
ā¢ methodologies
ā¢ technologies
4. Framework for the design & development
of Smart Environments
ā¢ accounting for technological and human / organisational aspects
ā¢ combining different dimensions and behaviour from
pervasive, distributed, situated and intelligent computing
Scope & Goals (2/2)
Butlers for Smart Spaces
ā¢ technology-neutral
ā¢ reference framework
ā¢ focused on usersā situated-
ness and interaction aspects
Home Manager
ā¢ platform for Smart Home &
Smart Living contexts
ā¢ focused on reasoning aspects
ā¢ multi-paradigm, agent-based
5. The Butlers Vision
ā¢ Butlers for Smart Spaces specialises the Butlers framework
to the Smart Spaces context
ā¢ Home Manager leads to concretise it as a multi-agent system
on the TuCSoN infrastructure
6. Butlers for Smart Spacesā¦
Specialisation of the Butlers framework to the Smart
Spaces context
The Monitoring layer groups together the
Butlers information and control layers
The Services layer embeds
the coordination referring to
the pre-processing of raw
information into exploitable
knowledge
Goals & Policies side-by-
side take into account user-
awareness (user-related,
higher-level coordination)
The Reasoning & Situated
Reasoning layers split the
Butlers Intelligence layer
7. ā¦on Home Manager as a MAS
Home Manager (apice.unibo.it/xwiki/bin/view/Products/HomeManager)
concretises it as a MAS on the TuCSoN infrastructure
The TuCSoN infrastructure conceptually
surrounds all layers, enabling and govern-
ing agent coordination & interaction
ā¢ All layers are re-shaped
based on TuCSoN concepts
& metaphors
ā¢ Agents & Policies sub-
layers appear side-by-side,
following the TuCSoN
approach
8. The Home Manager platform
ā¢ Open source platform for Smart Spaces, built on top of the
TuCSoN multi-agent infrastructure
o deployable also on a Raspberry PI 2
o Java-based
(~interoperable with Win10-IoT core)
ā¢ Smart House immersed in the surrounding environment
ļ Smart Living context
o Devices (air conditioners, lights, etc.)
o Users of different categories + RBAC
ā¢ Focus on Context Reasoning & Context Prediction
o Satisfy users desires while respecting global constraints
ļ suitable coordination laws to govern interaction
o Anticipate needs by exploiting the userās situation in time and space
9. Intuitive architecture
Main features:
o Autonomous āsituatedā decisions by exploiting the userās location
o Exploration of the environment around the userās location
o Information about the surrounding environment (e.g. weather)
o Interaction with selected social networks (e.g. Twitter)
o Tracking of the human presence
10. Smart Environments in HM
ā¢ Designing a Smart Environment in Home Manager amounts to:
o identify relevant device and service categories
o define a tuple-based representation of the relevant knowledge
o define the agent interaction
o develop an agent for each device category & service to interact with
Clear separation between
ā¢ social / individual intelligence
ā¢ mechanisms / policies
Features
ā¢ independent testing and debug
of agents and policies
ā¢ effective exploitation of the
data-driven, multi-paradigm
development approach
12. ā¢ Smart Fridge & Smart Pantry
o food monitoring
o collecting historical data on userās habits
o generate the corresponding buy tuple if necessary (policies)
ā¢ Smart Oven
o support the userās food cooking (e.g. dietary,ā¦)
ā¢ Smart Mixer
o recipe instructions interacting with Smart Fridge & Smart Oven to
check food availability ļ context adaptation
ā¢ Smart Shopper
o predict the userās needs ļ make contextualised suggestions
o shopping list based on the above data
o contact the āproperā vendor based on context- aware policies
The Smart Kitchen: scenario
13. The Smart Kitchen: prototype
Middleware ļ coordination laws ļ interoperability & integration
ā¢ Declarative approach
o bridge among different forms of heterogeneity
o support agent uncoupling
o support & promote separation between policies & mechanisms
o supports context reasoning
14. Future (current) work
Context prediction & adaptation: prediction agent
WEKA Classifier (Ilaria Bertoletti Master's thesis, 15 June 2017)
Tempera-
ture prefs
Device
usage
policies
Action plan on
air conditioners
Learning &
prediction of
user habits
15. Future (current) work
Context prediction & adaptation: prediction agent
WEKA Classifier (Ilaria Bertoletti Master's thesis, 15 June 2017)
CONTEXT
PREDICTION
ā¢ Grab user habits
info from multiple
sources
ā¢ Select relevant
data
ā¢ Anticipate user's
routine & desires
CONTEXT
REASONING
ā¢ Compute
heterogeneous
context info
ā¢ Enable HM to make
suggestions
ā¢ Increase HM
decision autonomy
PROACTIVE
ADAPTATION
ā¢ Autonomously appy
action plan on house
devices
ā¢ Enable HM to adapt
to new needs
ā¢ Reduce user's
interventions
17. Future (current) work
Context prediction & adaptation: prediction agent
WEKA Classifier (Ilaria Bertoletti Master's thesis, 15 June 2017)
18. Home Manager lays the foundations to support
context reasoning and context prediction
ā¢ Yet, just a starting point..
ā¢ A lot of work remains to be done
Conclusions (1)
Butlers for Smart Spaces
ā¢ technology-neutral
ā¢ reference framework for
pervasive IoT contexts
ā¢ focused on usersā situated-
ness and interaction aspects
Home Manager
ā¢ concretise the BSS approach
ā¢ the infrastructure bridges among
the agentsā ontologies, APIs,
knowledge representations,
interaction protocols
19. Conclusions (2)
Future work
ā¢ Deeper exploration of the context reasoning aspect
(machine learning,ā¦)
ā¢ Cross-platform interoperability
ā¢ Java/Windows 10 on the Raspberry
ā¢ Emerging standards
ā¢ Developing more complex policies and implementing other
advanced situated services
20. Home Manager URLs
ā¢ Home page
http://apice.unibo.it/xwiki/bin/view/Products/HomeManager
ā¢ Bitbucket repository
https://bitbucket.org/tuprologteam/homemanager