=Paper=
{{Paper
|id=Vol-1855/EUCognition_2016_Part8
|storemode=property
|title=Human-Aware Interaction: A Memory-inspired Artificial Cognitive Architecture
|pdfUrl=https://ceur-ws.org/Vol-1855/EUCognition_2016_Part8.pdf
|volume=Vol-1855
|authors=Roel Pieters,Mattia Racca,Andrea Veronese,Ville Kyrki
|dblpUrl=https://dblp.org/rec/conf/eucognition/PietersRVK16
}}
==Human-Aware Interaction: A Memory-inspired Artificial Cognitive Architecture==
https://ceur-ws.org/Vol-1855/EUCognition_2016_Part8.pdf
Human-Aware Interaction:
A Memory-inspired Artificial Cognitive Architecture
Roel Pieters1 , Mattia Racca1 , Andrea Veronese1 and Ville Kyrki1
Abstract— In this work we aim to develop a human-aware III. C OGNITIVE MODELING : MEMORY AND REASONING
cognitive architecture to support human-robot interaction.
Human-aware means that the robot needs to understand
The knowledge base is divided in declarative memory
the complete state of the human (physical, intentional and (semantic and episodic facts) and procedural memory (action
emotional) and interacts (actions and goals) in a human- library). Semantic facts is general knowledge to represent
cognitive way. This is motivated by the fact that a human the beliefs, relations and intentions of the world, of hu-
interacting with a robot tends to anthropomorphize the robotic mans and of objects. Episodic memory describes information
partner. That is, humans project a (cognitive, emotional) mind
to their interactive partner, and expect a human-like response.
about events and instances that occurred, e.g., what, where
Therefore, we intend to include procedural and declarative and when an event happened. The action library contains
memory, a knowledge base and reasoning (on knowledge base primitives and sequences of tasks available to the robot. For
and actions) into the artificial cognitive architecture. Evaluation example, the task model is encoded as declarative knowl-
of the architecture is planned with a Care-O-Bot 4. edge and describes the intention and relation between states
I. INTRODUCTION (phases) in a task. Moreover, it can also be described by an
action sequence and event sequence (episodic knowledge).
As the western world is aging, solutions have to be found
Reasoning over the knowledge base allows for fact checking,
that ensure the current high-quality welfare state for the
relation assessment and event comparison, and can be used
future. This research aims to assess the suitability of robotics
for future predictions (internal simulation). Reasoning over
for assistance and care. Such human-robot interaction should
the action library allows to reuse, adapt and augment actions
foremost be safe, intuitive and user-friendly. This implies that
and action sequences for different tasks.
the robot must understand the person’s tasks, intentions and
actions, and must include a knowledge base for information IV. S YMBOLIC TASK PLANNING AND EXECUTION
storage and reasoning. The main function of the symbolic task planner is to
II. P ERCEPTION : INTENTION AND TASK MODELING generate a suitable plan by checking if the task was ex-
perienced in the past (episodic memory in the knowledge
In order to provide assistance, the general state of the
base) and how (procedural memory in the action library).
human, as well as the task should be known. Human attention
Missing information for a generated plan is obtained from
can be used to understand a person’s intentions and the task
perception and reasoning over the knowledge base and the
he/she is engaged in. By detecting the head pose of the
action library. For example, actions take arguments that apply
human and projecting this into a 3D point cloud of the en-
to internal variables and functions (e.g., object pose, speech
vironment, a weighted attention map can be generated (Fig.
recognition). High level execution ensures that the planned
1-left). Segmenting this map returns the object of interest and
task is executed appropriately and the instructed goal is
can be used to determine which task the person is engaged in
achieved (Fig. 2).
[1]. Additionally, by actively gathering information (e.g., the
robot asking questions) a model of the task can be learned V. ROSE AND C ARE -O-B OT 4
(Fig. 1-right). This decision making problem under uncertain The proposed developments are part of the interdisci-
conditions can be modeled as a partially observable Markov plinary research project ROSE (Robots and the Future of
decision process (POMDP). By solving the POMDP, the Welfare Services2 ) which aims to study the social and psy-
robot can refine the task model, supervise the task execution chological aspect of service robotics. In particular, one aim of
and provide assistance for the next phase [2].
Fig. 1: Left: Weighted attention map that returns three objects
of interest, the plate received most interest (red). Right: Task
modeling scenario. A person is making a sandwich while a
NAO robot observes and asks questions to build a task model
Fig. 2: Artificial cognitive architecture for human-aware
for assistance.
interaction.
1 All authors are with School of Electrical Engineering, Aalto Univer-
sity, Finland. Corresponding author: roel.pieters@aalto.fi 2 http://roseproject.aalto.fi/en/
Proceedings of EUCognition 2016 - "Cognitive Robot Architectures" - CEUR-WS 38
this project is to investigate the requirements for social HRI
with elderly people and how these should be integrated in
practice. This applies for both the technological requirements
(i.e., what capabilities and algorithms are necessary) as well
as the social requirements (i.e., what does the user want).
The Care-O-Bot 4 will be used for human-robot interaction
studies and evaluation of the proposed artificial cognitive
architecture.
R EFERENCES
[1] A. Veronese, M. Racca, R. Pieters, and V. Kyrki, “Action and intention
recognition from head pose measurements,” 2017 (in preparation).
[2] M. Racca, R. Pieters, and V. Kyrki, “Active information gathering for
task modeling in hri,” 2017 (in preparation).
Proceedings of EUCognition 2016 - "Cognitive Robot Architectures" - CEUR-WS 39