PRESENTAZIONE

Cognitive Architectures for Robotics (COGAR) is a novel and highly experimental field involving research and development activities aimed at providing scholars and future researchers in robotics with conceptual tools and practical knowledge about how to design and implement complex architectures for robots involving advanced capabilities in perception, reasoning, and action. The subject starts by covers a few basics in software design, and goes on by investigating how to design and integrate perception, reasoning, and motion modules into a coherent workflow. Then, it showcases a number of proposed architectural designs in the literature, and features examples borrowed research-related use cases.

OBIETTIVI E CONTENUTI

OBIETTIVI FORMATIVI

Robots are multi-purpose, multi-form and multi-function machines. Next-generation robots are expected to exhibit completely new and unique behaviors with respect to current machines, specifically in terms of what they are meant to do, how they are structured, and what they are capable to do. In order to cope with this diversity in form and function, cognitive architectures for robots must designed to allow robots to perceive their environments, make sense of it, employing various knowledge representation and reasoning models, and then act effectively on their environment. The course will address all relevant topics in a research-oriented way.

OBIETTIVI FORMATIVI (DETTAGLIO) E RISULTATI DI APPRENDIMENTO

The main objective of Cognitive Architectures for Robotics (COGAR) is to provide students, scholars, and future researchers with advanced concepts and methodological tools about how to design advanced forms of intelligence embodied in robots and in the way they interact with the environments, the objects therein, and humans.

Robots are quickly becoming complex systems made up of modules and components whose behavior is complex in its own right, that is, not all the consequences of certain design choices can be easily predicted in advance to guarantee the overall robot behavior. The study and the synthesis of advanced forms of cognition and intelligence in robots requires:

./ defining which sensory information is needed, how it can be be processed, and how it can be integrated to form actionable information;
./ coupling sensory information and internal structures to represent information, in forms appropriate in terms of data recollection, integration, and synthesis of new knowledge;
./ designing algorithms to reason on such representation structures, for example to perform deduction, induction, or abduction of facts;
./ embedding knowledge representation structures and algorithms in software architectures designed to mimick advanced cognition capabilities, which may be inspired by those found in higher mammals or cephalopods, or based on completely different assumptions.

CogAR will provide a reasoned treatment of advanced concepts in cognition, state‐of‐the‐art design approaches, and advanced research trends in this field, as well as a comprehensive discussion about typical scenarios, solutions, and use cases.

PREREQUISITI

    Advanced knowledge of C/C++.
    Basic knowledge of Python and Java may be helpful.

MODALITA' DIDATTICHE

COGAR is organized into four key topics. Each key topic consists of theoretical concepts followed by practical work. Students are strongly encouraged to propose novel solutions to specific problems, which originate from real-world research challenges.

PROGRAMMA/CONTENUTO

Each year, the COGAR team updates the subject given new trends and development in cognitive robotics. This means that the subject is slightly different each year. COGAR introduces theoretical concepts and integrates them with practical insights. Examples and use cases are implemented using the widely adopted ROS framework. COGAR is based on the following macro-topics.

TOPIC 1: Introduction and motivations:

1/ Introduction to the course: what to expect, organization, exam.
2/ Cognition in humans and robots.

TOPIC 2: From software architectures to robot cognitive architectures:

1/ How we describe cognitive architectures: the Unified Modelling Language.
2/ Design patterns in Robotics for behavior-based robots, autonomous robots, and human-robot interaction.

Practice classes focused on TOPIC 2.

TOPIC 3: Classical robot architectures (with examples):

1/ Sense-Plan-Act architectures.
2/ Behavior-based architectures.
3/ Hybrid reactive-deliberative architectures.

Practice classes focused on TOPIC 3.

TOPIC 4: Cognitive architectures:

1/ Knowledge representation and reasoning techniques.
2/ Memory models, learning, and adaptation.
3/ Integrated task and motion planning.
4/ "Theory of Mind" and interaction.
5/ Consciousness and computation.

Practice classes focused on TOPIC 4.

TESTI/BIBLIOGRAFIA

Relevant material will be given by the teacher and the instructors during the lectures.

DOCENTI E COMMISSIONI

Commissione d'esame

FULVIO MASTROGIOVANNI (Presidente)

ENRICO SIMETTI

RENATO UGO RAFFAELE ZACCARIA (Presidente Supplente)

LEZIONI

INIZIO LEZIONI

https://corsi.unige.it/10635/p/studenti-orario

Orari delle lezioni

L'orario di questo insegnamento è consultabile all'indirizzo: Portale EasyAcademy

ESAMI

MODALITA' D'ESAME

The COGAR final mark is based on assignments. Assignments work as follows:

./ assignments are foreseen to evaluate in itinere how students, individually, understood the concepts and the methodologies after TOPIC 2, 3, and 4;
./ for TOPIC 2, the assignment consists of a multiple-choice exam on the matters discussed during classes;
./ for TOPIC 3, it consists of a programming-related assignment, in which students will have to design and develop a specific part of a robot architecture;
./ for TOPIC 4, it consists of a programming-related assignment, in which students will have to implement relevant traits of a cognitive architecture.

Please note that:

./ EMARO-wannabe and JEMARO students have a strict deadline to complete their assignment;
./ Ph.D. students attending the course can propose a topic on their own, agreed with us, as an assignment, and do not have any specific deadline;
./ the final grade will be a weighted mean of the grades of single assignments.

MODALITA' DI ACCERTAMENTO

The overall grade will be determined on the basis of the exhibited individual student's capability in applying the notions and insights discussed during the classes in the assignments.

Calendario appelli

ALTRE INFORMAZIONI

Follow-ups:

./ for successful assignments, we typically encourage students to co-author a scientific paper for Robotics-related conferences;
./ for EMARO-wannabe/JEMARO students, a few assignments can be continued as possible MSc thesis topics.