Robots as Intelligent Social Systems Working in the Outer World
Dubrovnik, Croatia, 17-22 June, 2013
Robotic International Summer-School 2013
Robots among us. Decisional and planning abilities by Dr. Rachid Alami, LAAS-CNRS, Toulouse, France
Human-robot interaction requires to equip the robot with explicit reasoning on the human and on its own capacities to achieve its tasks in a collaborative way with a human partner. This talk presents a robot control system which has been especially designed for a cognitive robot which shares space and task with a human. We have adopted a constructive approach based on effective individual and collaborative skills. The system is comprehensive since it aims at dealing with a complete set of abilities articulated so that the robot controller is effectively able to conduct a collaborative task with a human partner in a flexible manner.
These abilities include geometric reasoning and situation assessment based essentially on perspective-taking and affordances, management and exploitation of each agent (human and robot) knowledge in a separate cognitive model, human-aware task planning and human and robot interleaved plan achievement.
I will present and discuss these different topics and give concrete instances of their use.
Towards High Performance 24/7 Humanoids by Prof. Tamim Asfour, Karlsruhe Institute of Technology, Germany
Recently, considerable progress has been made towards the realization of humanoid robot systems which are able to move in a human-like way and perform tasks in human-centered environment. Although current systems are technologically advanced, they are still limited in their actuation, sensing, prediction, interaction and learning capabilities. Versatile humanoid robot systems integrating perception, action, prediction, planning and lifelong learning capabilities to carry out a variety of tasks in 24/7 manner in the real world are still missing.
In this talk, I will present recent progress towards building autonomous humanoid robots able to act, interact and perform complex grasping and manipulation tasks in a kitchen environment, to autonomously acquire object knowledge through active visual and haptic exploration and to learn actions from human observation and imitate them in goal-directed manner. The capabilities will be demonstrated on the humanoid robots ARMAR-IIIa and ARMAR-IIIb. The talk will emphasize scientific and technological bottlenecks and limitations of humanoid robot systems in general and discuss new research directions in grasping, action learning and the mechano-informatics of humanoid robots.
1. The role of force in the formalization and unification of sensorimotor representations. I will discuss how objects, agents and their actions can be described based on a new concept of sensorimotor force fields (SFF) which provides a unified representation and computational mechanism for solving tasks related to grasping, dexterous manipulation and robot control. SSFs are based on the assumption that different types of sensory modalities such as position, tactile, visual and even audio do share significant commonalities with each other as they can be mapped to force.
2. Semantic representations are a prerequisite for the development of cognitive capabilities and understanding in robots as well as for cooperation, interaction and communication with humans. Building such representations from sensorimotor experience rely on organizing the system’s sensorimotor experience to provide data structures which can be used at different levels of the systems hierarchy and breaks through the gap between sensorimotor level and symbolic level.
3. Humanoids for everyone. I will discuss the transformative impact of humanoid technologies on other fields and areas, where humanoid robots can be seen as wearable companions for augmentation or replacing of human performance in daily and working environment, in rehabilitation or in human-made and natural disaster. In particular I will address the mechano-informatics of humanoids.
Motion Coevolution of Robots with Overlapping Workspaces, by Dr. Petar Ćurković, University of Zagreb, Croatia
The talk is going to be about the problem of two industrial robots that work together to solve a complex assembly task. The robots are arranged so that they share workspace and thus present dynamic obstacle to each other. Simplified model of the two robots is implemented in Matlab and appropriate coevolutionary algorithm developed to optimize the movements of the robots so that they avoid collision and at the same time optimizes several additional criteria regarding the motions of the robots. The set of important criteria will be discussed, which yield motion results directly implementable from the simulation environment to the real robotic systems. The developed procedure may serve as a low-level routine on which different higher-level tasks may be reliably built.
Social Robotics meets Human-Robot Interaction: Developing Robots as Useful Assistants and Companions, by Prof. Dr. Kerstin Dautenhahn, University of Hertfordshire, United Kingdom
My lectures will introduce the research fields of social robotics in general and human-robot interaction (HRI) in particular. I will introduce some key concepts in these fields and highlight different viewpoints and approaches in HRI. Examples of companion robots that serve as assistive tools for elderly people or in therapy for children with autism will be presented. Such application areas in healthcare and therapy pose many challenges since they require a dedicated user-centred perspective as well as significant technological development and advances in machine learning and other fields. My talk will focus on scientific research challenges, open problems and methodological issues in such multidisciplinary research domains.
Knowledge Representation and Reasoning, by Prof. Alberto Fernandez, University Rey Juan Carlos, Madrid, Spain
Robots need to coordinate themselves in space, recognize and interact with the environment and the entities surrounding them. This requires efficient ways of representing information about the environment in which they are located. Robots need to perceive from the environment and create a mental representation of it. Also, robots must be endowed with cognitive knowledge to deal with the perceived information and decide which actions must be taken.
Furthermore, in open multiagent environments, after recognizing and “understanding” the surrounding World and its relations, robots need to exchange information/knowledge with others in order to achieve their objectives. Ontologies are key aspect of knowledge representation and reasoning.
This lecture will present basic concepts of knowledge representation using ontologies. In particular it will deal with three main aspects: (i) how to represent information using standard ontology languages (RDF, RDF Schema, OWL); (ii) what kind of automatic reasoning can be applied to information represented using ontologies; and (iii) how to consult information using ontology query languages (SPARQL). Finally, current trends to expose and access to machine-understandable linked data will be presented, which may be very appropriate in open dynamic systems. The lecture will include practical aspects, where pointers to ontology development tools and libraries will be given. Hand-on exercises on ontology building, reasoning and querying will be done during the sessions.
Robot team efficiency and task allocation, by Prof. Stefano Giordani, University of Rome “Tor Vergata”, Italy
In the real world robotic applications with scarce resources, optimization problems are essential for effective and efficient mission completion. At the theoretic level, they include multi-agent matching, min- cost, max-flow, shortest path, (robot) vehicle routing, etc. Usually, these problem require the integrality of the decision variables, namely, modeling through binary or integer variables. In this talk, a short presentation of some multi-robot team optimization problems will be given with the focus on the design and analysis of distributed and decentralized algorithms for the related integer optimization problems. The attention will be in particular devoted to the multi-robot task allocation problem. Theoretical results in the area of distributed and decentralized multi- robot negotiation through auctions, distributed Hungarian method, and other techniques as well as the application of our research to real-world robotic problems will be presented. The lecture is intended as a technical description of some state-of-the-art developments in advanced distributed and decentralized multi-robot optimization techniques, specifically the problems related with the assignment problem and distributed mission execution in the environments with limited and error-prone communication, with emphasis on mathematical theory, implementation, and practical applications.
Collective action by Dr. Andreas Herzig and Dr. Emiliano Lorini, Toulouse Institute of Computer Science Research, IRIT-CNRS, France
This talk presents basic concepts about individual and collective action. We will discuss some important logical theories developed in the last two decades to formalize individual and collective attitudes such as knowledge, belief (i.e. explicit and implicit belief), intention, group intentions, group beliefs (i.e. distributed belief, common belief, social acceptance). In the first part of the course we will focus on individual attitudes. In the second part of the course we will focus on collective attitudes.
Robotic challenges in new domains of application, by Prof. Bojan Jerbić, University of Zagreb, Croatia
Why do robots attract our attention? Probably, it comes from the ancient mystical belief that we are capable of designing own replica. There is also something that makes us terrified, as well as inspired in this sense. There is nothing more challenging then creation of new life, even artificial. But, the most of the robots do not resemble to humans, particularly industrial robots. Why are robots so impressive in the factory, but so incompetent in human environment? The everyday manipulation tasks we take for granted would puzzle the greatest robot bodies and brains in existence today. How to get to the new robotic era from the shared futuristic visions where robots become our ubiquitous partners that significantly affect our life? If we really want to, we need new robotic paradigms. Intelligent Robot should be a harmony that comes from the permeation of physical and virtual machines and environment concepts. Once the mind and the body were separated according to the Descartes Dualism, philosophers found it necessary to find a way to reconnect these two aspects. We need to find a way how the interaction between mind and body and the environment works as essential concept of human existence. Explaining how virtual and physical machines are related requires a deep analysis of causation. If we want to create really useful and intelligent robots we have to rethink the existing concepts within the background of new perceptions of culture, society, technology and ethic. The intelligent robots should possess machine consciousness and capability to understand the context of its world. Then we could expect really influential application in the domains other than industrial (medicine, home etc.)
Reusable Teamwork for Multi-Robot Teams, by Prof. Gal A. Kaminka, Bar Ilan University, Tel-Aviv, Israel
For many years, multi-robot researchers have focused on specific application-inspired basic tasks (e.g., coverage, moving in
Human civilization as technology based civilization, by Prof. Branko Katalinić, Vienna University of Technology, TU Vienna, Austria
Our civilization advanced from food-gathering tribes to Internet supported commercial-based urban civilization of today. Human civilization has become technology-based civilization. Science and technology have been two main strategies of success of Homo Sapiens. Some of the main human characteristics are the abilities to think, plan and design. Starting from innate human capabilities, we need to find out the actual development line. Main mega trends in science and technology determine our future. We have begun to focus the scientific research on the relationship between humans and their artifacts, emphasizing the importance of the use of intelligence, tools, interactions with machines, and languages for the development of technology. Engineers therefore become increasingly important factors responsible of our future. Doctor of technical sciences as the highest academic level of engineering education can be described by about twenty characteristics, implying the development of engineering as well as the development of our society as such. The traditional views of education and technology of the western society need to be revised so as to be able to envision new future scenarios of the development of technology with all the challenges, risks and opportunities tackled on the way.
Autonomous knowledge acquisition through interaction with environment by Prof. Danica Kragić, Royal Institute of Technology, KTH, Stockholm, Sweden
The ability to autonomously acquire new knowledge through interaction with the environment is one of the major research goals in the field of robotics. The knowledge can be acquired only if suitable perception-action capabilities are present. In other words, a robotic system has to be able to detect, attend to and manipulate objects in the environment. In the first part of the talk, we present the results of our longterm work in the area of vision based sensing and control. The work on finding, attending, recognizing and manipulating objects in domestic environments is discussed. More precisely, we present a stereo based active vision system framework where aspects of Top-down and Bottom-up attention and foveated attention are put into focus and demonstrate how the system can be utilized for object grasping.
The second part of the talk presents our work on the visual analysis of human manipulation actions which are of interest for e.g. human-robot interaction applications where a robot learns how to perform a task by watching a human. A method for classifying manipulation actions in the context of the objects manipulated, and classifying objects in the context of the actions used to manipulate them is presented. The action-object correlation over time is then modeled using conditional random fields. Experimental comparison shows improvement in classification rate when the action-object correlation is taken into account, compared to separate classification of manipulation actions and manipulated objects.
Self-* Multi-agent Systems in Robotics, by Prof. Paulo Leitao, Polytechnic Institute of Bragança, Portugal
The lecture discusses the application of artificial intelligence techniques, and particularly multi-agent systems, to develop intelligent, complex and adaptive engineering systems, as manufacturing and robotics are. The enrichment of such systems with mechanisms inspired in biology, e.g. self-organization, can greatly contribute to increase the system performance, flexibility, robustness, adaptation and re-configurability, taking into consideration that in nature very complex and adaptive systems are implemented by using very simple behaviours.
Initially, the theoretical aspects of engineering such emergent and adaptive multi-agent systems will be presented. A special attention will be devoted to analyse the integration of self-* properties, such as self-organization, self-adaptation and self-learning, in multi-agent systems. To conclude, some applications of these concepts to robotics will be discussed, e.g., in surveillance, demining and soccer.
Robot team missions with incomplete situational awareness, by Dr. Marin Lujak, University Rey Juan Carlos, Madrid, Spain
In this lecture the problem of a decentralized target allocation for a team of mobile robots (UAVs) will be investigated. The success of the multi-robot team task allocation is highly influenced by the team cooperation and efficiency of resource utilization. Efficient group's coordination results in the minimization of the mission's total cost (e.g., fuel and other supplies consumption, crew size, and mission duration), while in the case of adversarial environments, it also results in the maximization of adversarial neutralization with minimal own-group damage. The lack of cooperation of one or more agents within the group due to the lack of agreement or situational awareness gives suboptimal mission achievements and in some cases can be fatal. We will tackle the issues of incomplete situational awareness and how the efficiency of mission accomplishment is influenced by the lack of updated information in dynamically changing environments. The emphasis will be on the strategic behaviors and algorithms for task assignment capable of dealing with incomplete information.
Engineering open hybrid social environments through normative systems, by Prof. Pablo Noriega, Artificial Intelligence Research Institute IIIA-SCIS, Barcelona, Spain
Over the past decade we have witnessed a growing number of web-enabled practices that involve human as well as non-human "rational" agents who interact in some sort of coordinated way. Such practices may range from environments to support social play among disabled children to user-directed efficient utility grid management. Because of the complexity and liability involved in many of these practices, it is desirable to use a principled way in their design and implementation. In this seminar we shall explore that possibility. We will characterize the class of "open hybrid multiagent environments", we will discuss the requirements for their proper engineering and show one way of implementing them using "electronic institutions". Then we shall argue for generalizing this particular approach with the help of "normative" notions and suggest some open problems that may be of particular interest for social intelligence, in general, and robotics research in particular.
Talks abstracts (in lecturers’ alphabetical order)