The Cognitive Robotic Systems Laboratory at AASS
(Previously "Mobile Robotics Lab")
Our Research Themes
Please note: This page has not been updated in a
while. If you are interested in the current research thems of
our lab, you can either wait that we all get a bit of quiet time and
update it, or contact directly the Lab
Leader.
 |
The scope of our investigation is not limited to traditional robots, but
more generally to any intelligent system that is physically embedded in
the environment through sensors and actuator. Mobile robots are
instances of such systems, but simpler devices like intelligent home
appliances or devices to interface with a human user also fit into this
category.
|
|
We schematize a Physically Embedded Intelligent System (PEIS)
as shown on the right. A system of this kind includes both cognitive
functionalities that enable abstract reasoning (Modeling and
Deliberation), and sensori-motoric functionalities that enable
situatedness in the physical world (Perception and Control). Two
assumptions underlie our otherwise general schema:
- That the system is modular, that is, it is composed of a number
of individual functional or behavioral components.
- That the system is layered, that is, it consists of a cognitive
layer and of a sensori-motoric layer.
|
|
We take inspiration from the fields of artificial intelligence and of
robotics to realize the above two layers, respectively. Our aim,
however, is not to perform incremental research in these two fields
per-se, but to use knowledge and techniques from those fields to address
the integration problem for a physically embedded intelligent
system.
|
|
We address several facets of the integration problem. First, we study
the integration within a single PEIS -- namely, between the
high-level cognitive functionalities (M, D) and the low-level
sensori-motoric ones (P, C). Second, we study the integration
across multiple PEIS -- how to make multiple PEIS cooperate
during the performance of tasks. Finally, we study the integration of
(single and multiple) PEIS with the environment, e.g. using
non-standard sensor modalities like olfaction, and with the
humans who inhabit the environment.
|
|
The above facets are reflected in our four broad research themes:
Cognitive robots,
Robot ecologies,
Artificial olfaction, and
Robots for the humans.
|
|
Theme 1: Cognitive Robots
|
|
A physically embedded intelligent system reasons about the pysical world
which it is embedded in. Deliberation is needed to connect its
contingent actions to its global goals and long-term desires. To be
effective, the knowledge on which deliberations are based must be
consistent with the physical world; symmetrically, the results of
deliberation must be correctly translated into physical actions. In
other words, the system must make sure that its conitive processes are
in sync with the physical world.
The general question addressed in this research theme is how the above
syncronization can be achieved. This question is central to our ability
to realize cognitive robots, that is, systems that can both interact
with the physical world and reason about it. In our generic PEIS agent,
this question translates in how to realize the links between the
cognitive functionalities (M+D) and the sensori-motoric functionalities
(P+C).
|
|
The first part of the above question is how to connect knowledge-based
models (M) with perception (P). Symbolic names, like cup-22,
are used to denote objects in most AI reasoning and planning systems.
When the robot must physically access these objects, however, it must
rely on the data provided by its sensors. For instance, to perform the
action Pickup(cup-22), a robot must use the data provided by
its camera about the position and shape of the cup. We call
perceptual anchoring the problem to create, and to maintain in
time, the connection between symbols and sensor data that refer to the
same physical objects. Perceptual anchoring is closely related to the
symbol grounding problem, one of the fundamental problems in artificial
intelligence, but it addresses a specific case: one in which the symbols
refer to physical objects. Our objective within this research theme is
to provide practical solutions to the anchoring problem, despite of its
philosophical complexities. To be practical, these solutions must also
adress problems like observation failures, sensor uncertainty, and
perceptual ambiguity.
|
|
The second part of the above question is how to connect goal-oriented
deliberation (D) with control and reactive execution (C). The field of
AI has witnessed impressive development in the design of knowledge-based
planners able to deal with complex tasks in an efficient way. The
generated plans often assume that the executing system can reliably
execute abstract actions, and that it can correctly estimate the current
state of the world. Physical perception and execution, however, are
inherently uncertain and may fail, especially in a dynamic world. Our
objective within this research theme is to develop planning techniques
that can be used to guide reactive behavior in the presence of
uncertainty and dynamicity. The generated plans must include provisions
for reacting to new perceived information, as well as for acquiring this
information when needed. An interesting case is when dynamicity and
uncertainty are introduced by the presence of humans in the environment
where the robot operates.
|
|
This research theme is currently reflected in the following concrete projects:
|
|
In addition to studying physically embedded intelligent systems in
isolation, we are interested in studying collections of such systems.
In this research theme, we take an ecological viewpoint in which the
robots and the environment are seen as parts of the same system. Robots
can be pervasively distributed throughout the environment in the form of
mobile platforms, embedded sensors, actuators, or smart objects, and
they can engage in symbiotic relationships.
This research theme is devoted to study how robot ecologies of this type
can collectively perform tasks. The general paradigm that we adopt is
that each physically embedded intelligent system in the ecology can
"borrow" functionalities from other systems in order to compensate or
complement its own. Alternatively, one could say that we consider an
overall physically embedded intelligent system whose components are
distributed among a set of separated physical entities.
|
|
This research theme naturally interacts with the other themes. In
particular, we are interested to investigate how a cognitive robot can be realized in a distributed
way, and how to deal with perceptual anchoring and of plan execution in
the distributed case. We also investigate robot ecologies capable of olfaction and robot ecologies that include humans, especially in the context of elderly care in
domestic environments.
|
|
This research theme is currently reflected in the following concrete projects:
|
|
Theme 3: Artificial Olfaction
|
|
We have a special interest in a sensor modality which has been scarsely
used in robotic applications until now: olfaction. In this research
line we study techniques for artificial olfaction, and their integration
within an intelligent robotic system. This research line has strong
interactions with the other lines: these include the integration of
olfaction with knowledge and cognition, the
enrichment of a robot ecology with olfaction, and
the use of olfaction in medical applications.
|
|
This research theme is currently reflected in the following concrete projects:
|
|
Theme 4: Robots for Humans
|
|
In this research theme, we investigate the use of robotic technologies
to improve the quality of life of humans. We are especially interested
in the applications of robotic technologies to health care, with special
emphasis on the elderly population. Robotic technologies are intended
here in a broad sense, including for instance sensing and sensor
interpretation. Our approach in this research theme is highly
interdisciplinary: we work closely with researchers from other relevant
disciplines, like medicine, psychology, and nursery; with social actors
involved in organizing or delivering elderly care; with economic actors
involved in the creation of relevant products; and with end users like
patients or elderly people.
|
|
This research theme is currently reflected in the following concrete projects:
|
|
