IURS06

6th International UJI Robotics School IURS 2006

 

Humanoid Robots

 

September, 18-22, 2006

Bonaire Hotel, Benicasim, Spain

Introduction

This year the Robotic Intelligence Lab of Jaume-I University, has organized a major event on the research and development of Humanoid Robots, in collaboration with the Universität Karlsruhe (TH), and EURON, the European Robotics Research Network.

The school lecturers are leading world experts in the building and development of Humanoid Robots and related skills.

 

Aims and Scope 

The attention of the robotics community has been drawn more and more on humanoid robots in the last years. This interest is not only motivated by the trend of designing robots with human appearance, but also for the implications of their use in human environments. They must be able to perform a wide range of different tasks in partially or completely unknown environments. And, what is most interesting, they must able to cooperate and probably communicate with humans in a variety of modes. The development of such amount of different capabilities represents and ambitious and attractive research field for many scientists.

First of all, we focus on the grand challenge that the development of human-like robots represents for mechatronics engineers. The mechanical design of a humanoid robot must be anthropomorphic not only in their appearance but also in their capabilities. And this often implies that the different parts must be light-weighted and highly versatile, that is, with a high number of degrees of freedom. In addition, the development of humanoid robots results in the research fields that otherwise would have a smaller area of application. An example of these, is the development of anthropomorphic arm and hands, the development of stereo heads, and the research on biped robots.

Probably the most exciting interest of humanoid robots is their intense interaction with humans and their appropriateness for tasks in human-centered environments, due both to their friendly appearance an their anthropomorphic design. But the exploitation of these capabilities requires the development of novel control strategies, and, more interestingly, more advanced human-robot cooperation and communication skills. Examples of the last are learning by imitation, language acquisition, and gesture recognition, among others.

The purpose of this school is to give the students a general background on these main topics. To achieve this goal, world-wide known lecturers would introduce the students in the basic concepts of humanoid design and programming. The lecturers have been selected not only because of their knowledge in particular topics related with humanoid robots, but also for their valuable experience in the building and programming of real humanoid robots, and by their outstanding teaching experience.

Summarizing, the topics to be addressed in the program during the summer school are grouped in the following items:

  • Humanoid robot building:
    • Anthropomorphic arm/hands
    • Biped locomotion
    • Robot heads
  • Advanced skills
    • Dexterous manipulation
    • Advanced stereo vision
    • Skill learning
  • Human-robot cooperation
    • Learning by demonstration
    • Language acquisition
    • Gestural communication

The participants will get in touch with other young researchers in the field and will have the opportunity to improve their knowledge through the interaction with experts. Another important objective is to provide European graduate students in robotics with a unique training opportunity in this emergent and fast-evolving domain. Students will closely interact for one week with some of the top-level researchers in the world in this domain. They will also have the opportunity to discuss their research work with them, as well as with other European students.

 

Lecturers 

Rüdiger Dillmann

Institute of Computer Science and Engineering
Industrial Applications of Informatics. & Microsystems
Universität Karlsruhe (TH), Germany
> <a href="http:/wwwiaim.ira.uka.de">http://wwwiaim.ira.uka.de

Carlos Balaguer

RoboticsLab, Universidad Carlos III, Spain
http://roboticslab.uc3m.es

Roderic Grupen

Laboratory for Perceptual Robotics
University of massachussetts, USA
http://www-robotics.cs.umass.edu/

Ales Ude

Department of Automatics, Biocybernetics and Robotics
Jožef Stefan Institute, Slovenja
http://www.cns.atr.jp/~aude/

Cecilia Laschi

Advanced Robotics Technology and Systems Laboratory
Scuola Superiore Sant'Anna, Italy
http://www-arts.sssup.it/people/assistantprofessors/claschi/

Tamim Asfour

Institute of Computer Science and Engineering
Industrial Applications of Informatics. & Microsystems
Universität Karslruhe (TH), Germany
http://wwwiaim.ira.uka.de

Luc Berthouze

Neuroscience Research Institute (AIST)
Tsukuba Central 2, Japan
http://staff.aist.go.jp/luc.berthouze/

Angel P. del Pobil

Robotic Intelligence Laboratory,
Universitat Jaume I, Spain
http://www3.uji.es/~pobil

Martín Mellado

Instituto de Automática e Informatica Industrial, ai2
Universitat Politécnica de Valencia, Spain
http://www.ai2.upv.es

 

Titles and short abstracts of the lectures

Lectures will be given in English. The instruction will take place for a period of one week from Monday to Friday (Sept 18-22, 2006). Preliminary schedule is from 9:00 AM to 12:00 noon and from 2:00 PM to 5:00 PM, with one afternoon free.

 

Rüdiger Dillmann

Emergent Cognitive Capabilities of Humanoid Robots

Humanoid robot systems should have the ability to interact with humans in terms of conversation about the task to be done and how to be done the task and finally to execute it goal-oriented and to react on disturbances or unexpected events in a competent manner. Such robot systems should be able to operate in dynamic human centred scenarios which require abilities such as adaptivity, perception and learning. Examples for such systems are service robots or humanoids that interact in the immediate environment of humans in a context dependant goal-oriented manner and cooperate with humans. The behaviour of such robots is characterized by active sensing processes, fusion of sensor data, perception and the selection of appropriate actions as well as their superimposed control system. Thereby learning and shaping of sensory and motor abilities as well as the active observation and interpretation of situations and actions are of major interest.

A probate approach for learning knowledge about actions and sensory-motor abilities is to acquire prototypes of human actions by observation the actions with sensors and transfer these abilities in the sense of learning by demonstration to the robot. This requires human motion capture, observation of interaction, object state transitions and observation of spatial and physical relations between objects. By doing this, it is possible to acquire so-called "skills", situative knowledge as well as task knowledge, and can be introduced to new and unknown tasks. New terms, new objects and situations, even new types of motion can be learned with the help of a human tutor or be corrected interactively via multimodal channels. The term multimodality describes communication channels which are intuitive for humans, such as language, gesture and haptics (physical human-robot contact). These are to be used for commanding and instructing the robot system.

The field of programming by demonstration has been evolved strongly as a response to the needs of generating flexible programs for service robots and is largely driven by attempts of modelling human behaviour and to map it onto virtual Androids or humanoid robots. It comprises a broad set of observation techniques processing large sets of data from high speed camera systems, laser, data gloves and even exoskeleton devices. Some operate with precise a- priori models other use statistical approaches to approximate human behaviour. In any case observation is done to identify motion over space and time, interaction with the environment and is effects, useful regularities or structures and is interpretation in a given context. With this goal systems have been developed which combine active sensing, computational learning techniques, multimodal dialogues to enrich the semantic system level, memorisation techniques as well as mapping strategies to make use of the learned knowledge to control a real robot.

 

Carlos Balaguer

Why humanoid robots?

The tremendous advances of the Japanese humanoids during the last years make that robotics community actively debates the role of the humanoids in the future robotic society. Questions like “legged or wheeled?”, “active or passive control?”, “full size or toy humanoids?”, “cost or performance?”, “what type of applications?”, “rich or low on-board perception?”, “active cooperation with humans?”, “one hundred percent security?”, “how to increase the power autonomy?” (etc.) are examples of that debate. The talk will introduce the main advantages of the humanoids robots together with the open issues and will try to analyze (and/or answer) the mentioned questions through illustrative examples.

 

Humanoid robots’ gait control strategy based on the Lie logic technique and LIPM model

The gait control is one of the most important issues of the legged humanoid robots. The talk will introduce the stable walking through COM and ZMP concepts, in both static and dynamic domains. The kinematics transformations of high DOF legged humanoids have been solved nowadays using the Lie logic techniques, canonical Paden-Kahan sub-problems and POE formulation, concepts that will briefly introduced. To simplify the computational implementation, the gait algorithms are commonly based on 3D-LIPM (Linear Inverted Pendulum Model) in order to get natural and dynamic biped walking. The described techniques and algorithms will apply to the legged Rh-1 full size 21 DOF humanoid robot.

 

Roderic Grupen

The Developmental Organization of Manual Dexterity for Humanoid Robots

Human infants display a tremendous assortment of time-varying structure in their physiological and neurological responses to the world. We propose that kinematic, dynamic, and neurological properties of a developing infant are exploited to simplify and structure learning in the context of an on-going interaction with the world. Developmental processes construct increasingly complex mental representations from a sequence of tractable incremental learning tasks. Our goal is to provide computational mechanisms for modeling these aspects of developmental in order to program complex robot systems.

We will focus on the development of reaching, grasping and dexterous manipulation in bimanual humanoid robots. We start from a relatively optimistic position, namely that traditions in robotics, control theory, AI, and learning are adequate computational accounts of some critical aspects of developing human infants and present important concepts illustrated with examples on Dexter---a bimanual humanoid. We will discuss:

  • Kinematic, Dynamic, and Maturational Structure
  • Dynamical Systems Approach
  • Neurological Structure - Developmental Reflexes and Composability
  • A Developmental Assembler
  • Schemata and Programming by Demonstration

 

Ales Ude

Humanoid robot vision (or Foveated vision for humanoid robots)

A robot vision system can be called humanoid if it firstly possesses an oculomotor system similar to human eyes, and secondly if it is capable of simultaneously acquiring and processing images of varying resolution. Different possibilities for the technical realization of such visual systems will be reviewed. Through a case study of object recognition on humanoid robots, some motor control algorithms and visual processes that can make use of such systems will be presented.

 

Generation of full-body movements for humanoid robots by observation

The formulation and optimization of joint trajectories for humanoid robots is quite different from this same task for standard robots because of the complexity of humanoid robots’ kinematics and dynamics. This lecture will focus on how to exploit the similarity between human motion and humanoid robot motion to generate joint trajectories for humanoids. Automatic approaches to relate humanoid robot kinematic parameters to the kinematic parameters of a human demonstrator will be presented.

 

Cecilia Laschi

Bioinspired tactile perception

This talk has the objective of providing the attendees with an analysis of the human tactile system, as a model for designing robot tactile sensors, and with basic knowledge on the fundamental technologies for developing tactile sensors for robots.

The talk will start with a description of the human sense of touch from an engineering point of view, including characterizations of tactile receptors and of the fingertip, seen as a tactile organ, with its role in grasping. The main technologies for developing tactile sensors will be then analysed, by explaining the underlying working principle, with mathematical relations, and by reviewing state-of-the-art case studies.

 

Vision and eye movements in Humans and robots

This talk has the objective of teaching the principles of active vision in humans and robots. A review of the fundamental neurophysiological principles of human vision, including the eye movements that are strictly related with visual perception and visuo-motor tasks, will be given in the first part of the talk. A possible approach to develop retina-like vision in robots and to develop human-like robotic heads reproducing eye movements will be then described, by underlying the coordination between the perceptual and motor tasks.

 

Tamim Asfour

Humanoid Robots: Design Issues and Control

The design of humanoid robots requires coordinated and integrated research efforts that span a wide range of disciplines.

The talk will focus on the design issues of the upper body of humanoid robots. Through a case study of the humanoid robot ARMAR-III, the kinematics modelling, motion control will be presented. Furthermore, the hardware and software architecture which allows the integration of perception and action will be demonstrated.

 

Imitation Learning of Human Arm Movements

Robot learning by imitation, also referred to as programming by demonstration has been dealt with in the literature as a promising way to teach humanoid robots and several imitation learning systems and architectures based on the perception and analysis of human demonstrations have been proposed.

This talk deals with imitation learning of arm movements in humanoid robots. Different imitation architectures will be presented. A possible approach that makes use of Hidden Markov Models (HMM) to learn and reproduce movements demonstrated by a human instructor multiple times will be described.

 

Luc Berthouze

Motor development in humanoids.

Taking inspiration from developmental psychology and biology to develop humanoids that acquire motor skills. Case studies: swinging, bouncing. The lecture will deal with important issues for complex motor systems: the degrees of freedom problem (I will talk about freezing and freeing of the degrees of freedom as a way to simplify the acquisition of coordinated control), entrainment, a property of oscillatory control structure which provides great flexibility in control against perturbations, and changes in the system (morphological changes for example).

 

Angel P. del Pobil

Some critical -and unsolved- problems in Humanoid Robotics and how learning and biological inspiration can help solve them

 

Martin Mellado

Practical hints about laboratory activities

 

Other activities

There will be one whole day devoted to practical exercises in the lab . Other activities will include:

  • Active siesta. In order to facilitate the interaction between students and senior scientists a so-called active siesta will be organized everyday, for the students to informally interact with the lecturers. Speakers will be asked to be available after lunch from 1:30 to 2:00 p.m. for informal discussion with students. Each student can choose a discussion group moderated by one of the instructors who typically meet outdoors on a cup of coffee, taking advantage of the hotel outdoor garden and good weather.
  • Students presentations. In order to maximize interaction among students, they are invited to briefly present their research work. For this purpose a 30-min slot is reserved everyday in the timetable right before the first morning lecture.
  • Students posters. Also, students are invited to bring their own posters so that after their talks they can easily have further discussions with those colleagues that were working on related topics. A specific poster session is scheduled.

 

Laboratory activities

Laboratory sessions will be organized, where the students will implement some of the techniques they have learnt. In the labs, students will be organized in groups, encouraging interactive work among them.

A walking robot platform, still to be decided, will be used during a whole day practice. The different groups will develop some walking programs for the robot and a final competition will allow the comparison, in an informal way, of the throughput of their proposals.

The laboratory activities will be organized in collaboration with Dr Martin Mellado from Polytechnical University of Valencia (UPV).

 

Organization 

General Chair: Program Chairs:    
Angel P. del Pobil
Universitat Jaume I, Spain
Organization Chair
Antonio Morales
Universitat Jaume I, Spain
Rüdiger Dillmann
Universitat Karslruhe, Germany
Pedro J. Sanz
Universitat Jaume I, Spain

Local Organization:

  • Antonio Morales, Organization Chair
  • Martín Mellado, Laboratory activities
  • Ester Martinez Martín
  • Miriam Martinez Martín
  • Mar Marcos López
  • Begoña Martinez Salvador
  • Patricio Nebot Roglá
  • Mario Prats Sánchez
  • Raúl Wirz
  • Pablo Dominguez Romo
  • Raúl Marín Prades
  • Enric Cervera Mateu
  • Gabriel Recatalá Ballester