Modeling of ADLs in its Environment for Cognitive Assistance
Jérémy Bauchet and André Mayers
Introduction
Cognitive assistance, in smart homes, aims at supporting occupants for the completion of their activities of daily living (ADLs)
Introduction (2)
Implies for the system a prior knowledge about the occupant its activities its environment
This knowledge is necessary for :activity recognition, as a prior step of cognitive
assistance in smart homes finally, for cognitive assistance
Plan
Introduction A model for the description of ADLs
a hierarchical approach the environment of completion
Taking into account the specific behavior of the occupant Implementation of the models Results and perspectives concerning activity recognition
and cognitive assistance Conclusion
A hierarchical model for ADLs description Two type of nodes : tasks and methods
a task : a goala method : a way to realize the task
→ a set of subtasks
→ and rules of integration of subtasks : partial or total sequence repetition or necessity constraint
A hierarchical model for ADLs description (2) Roots are abstract tasks (ADLs, IADLs) Leaves are methods of terminal tasks = an
atomic way to realise a concrete goal
Tasks can be common to several methods→ if common nodes are duplicate, this
model is a tree
Model of activity
A model of ADLsin its environment Environment of activity completion Includes all actors of activity completion
daily living objects furniture the occupant, as the actor of his own task
completion
→ e.g. : her/his current position
Description of the environmentof completion Static description :
Actors : fridge Events concerning actors
fridge : door opened, door closed
Dynamic description : Assertions, giving current value of several pieces of information concerning actors
<fridge, door, opened> <occupant, position, kitchen>
Links between activities andthe environment Events concerning actors in the environment
are associated with terminal methods events are a consequence of the concrete actions
of the occupant events can be observed via distributed sensors occurrence of events are used for activity
recognition
Links between activities andthe environment (2) Tasks and method are considered as
operators of a planning domain they have preconditions and effectsboth concern the environment
Taking into account the specific behavior of the occupant The activity model is a support for the
generic description of ADLs and IADLs
→ We need an occupant model to describe his/her specific comportment
An episodic memory for the occupant model Allows to precise how one occupant
usually completes an activity the method used for a given task the time slot of completion the location the sequence of subtasks
Implementation
XML library of tasks and methods (activity model) description of the environment episodic memory persistence
SVG graphical representation of the environment
Java XML parsing internal representation of the models and treatments SVG management
Results and perspectives : Concerning activity recognition Goal : to compute the probability of
completion of (I)ADLs given inputs Inputs are :
description of ADLseventscurrent timeknowledge about the occupant habits of life,
given by the episodic memory
Activity recognition
Description of activities, for step by step or global assistance
Preconditions → what has to be done before, where the activity can take place
Rules of integration for subtasks → how to complete correctly the activity
Episodic memory → anticipation process
Results and perspectives (3) :Concerning cognitive assistance
Conclusion