URL
http://www.cs.bham.ac.uk/research/projects/cosy/matrix/active-animate-spatial-general/axs-fido-animate-active.html
Last changed: 6 Dec 2005
Animate Active objects -- a General Overview
Targets for the long term future (FIDO)
(DRAFT -- liable to change.)
Another
file introduces a general category of active objects, that Fido, the
domestic robot of the future will need to know about and deal with.
Here we consider the subset of 'animate' active objects, namely those
that are
not passive
(they have their own store of energy, or the
ability to re-direct external sources of energy), they are
self-controlled
insofar as they select uses of their energy on the basis of the states
of various internal and external sensors (either using raw readings the
outputs of sophisticated interpretation processes), and they
can generate goals
which determine how they act in combination with
sensor values -- i.e. goal states and internally and externally sensed
information affect
the deployment of energy, i.e. they
-
can take internal or external actions
-
they select and modulate actions on the basis of
-
current sensed internal and external state
-
current goals
This charactarisation would be application to a wide range of animals
and simple machines. For example, Fido may need to interact with and
possibly communicate with various humans, e.g. not only its
owner/employer but other members of the family (possibly including
children who need looking after), visitors, doctors, tradesmen, postmen,
and other service providers, in addition to pets of various sorts, and
possibly also machines of varying degrees of intelligence, including
possibly one or more other robots with similar capabilities to Fido.
There are enormous variations in complexity and sophistication of these
animate participants in Fido's life, including
-
what sorts of sensors are available,
-
how the sensory information is processed,
-
how the intermediate stages and the results
of processing are represented,
-
what kind of ontology is used
-
what the various kinds of information are used for and how they are used
-
what sorts of short term, medium term and long term information
storage and retrieval mechanisms are available and what they can do
(e.g. can they store hypotheses, unanswered questions, generalisations,
explanatory theories, plans, ...?)
-
what kinds of architectural components are present, e.g.
-
whether all behaviours are purely reactive
-
whether there are deliberative capabilities (e.g. with the ability to
construct, compare and evaluate representations of possible future,
past, distant or other unperceived states and processes, and to
take decisions and produce actions on the basis of all that,
-
whether there are internatl self-monitoring capabilities and if so what
sorts and what ontologies do they use
-
how far the architecture is fixed from the start or whether there are
mechanisms for modifying and extending the architecture
-
whether there are personalities that can change and develop over time
-
etc.
In principle a domestic robot might be needed to know a huge amount
about human life and society. But we could for present purposes restrict
its capabilities to those of a child of about 5 years at most, and its
functions to those that a child helper could provide in the household,
though the robot may have much more strength than the child and may have
some specialised functions that go well beyond a child, e.g. special
planning capabilities for dealing with medical or other emergencies
involving the person it is helping or looking after.
One of the difficult tasks in generatng requirements for a long term
domestic robot development project is specifying, or more precisely
delimiting, the required extent of understanding of human minds -- how
they perceive things, what they want or need, what they know or forget,
what their likes and preferences are and how their moods can change, how
the preferences and goals can change over time, how they learn, what
sorts of help they want or do not want, how things can go wrong when
they act while their capabilities are impaired.
Probably for present purposes it may be best to restrict consideration
to
-
robot tasks involving meeting physical needs of humans, including
-
handling medicines and helping the owner take them
-
providing food,
-
dealing with objects connected with eating including
chairs, tables, cutlery, glasses, bottles, and other containers,
the food itself,
and
the devices that are associated with food preparation and cleaning
up after meals
-
storing, fetching and manipulating other objects such as books,
magazines, writing implements, keyboards, TV controls, etc.
-
helping with dressing or undressing, pushing a wheel
chair if necessary, fetching or delivering objects of various sorts from
or to parts of the house, dealing with callers who ring the doorbell,
-
cleaning and tidying the house
and also
-
obtaining and providing information needed by humans at various times
such as telling a person where things are, or what is happening out of
sight, or what the plans for the day are, etc.
Precisely what sort of information and what help with dealing with
information is needed will depend, for example, on whether the person
being helped is blind or not, deaf or not, intellectually impaired or
not, etc.
See discussion of shorter term targets for end of cosy project
here.