TechTalks from event: ICRA 2013 Plenary Talks

  • RoboBees: progress in insect-scale robotics Authors: Robert Wood, Harvard University
    As the characteristic size of a flying robot decreases, the challenges for successful flight revert to basic questions of fabrication, actuation, fluid mechanics, stabilization, and power - whereas such questions have in general been answered for larger aircraft. When developing a flying robot on the scale of a common housefly, all hardware must be developed from scratch as there is nothing “off-the-shelf” which can be used for mechanisms, sensors, or computation that would satisfy the extreme mass and power limitations. This technology void also applies to techniques available for fabrication and assembly of the aeromechanical components: the scale and complexity of the mechanical features requires new ways to design and prototype at scales between macro and MEMS, but with rich topologies and material choices one would expect in designing human-scale vehicles. With these challenges in mind, this talk will present progress in the essential technologies for insect-scale robots and the first flight experiments with robotic insects.
  • Multi- and Cross-Lingual Robotic Assistants Authors: Alexander Waibel, CMU and KIT
    As humanoid robotics is coming of age and a growing number of mechanical functions become possible, a robot’s perceptual and cognitive functions continue to grow in importance and R&D focus. While perceptual abilities of machines generally still lag behind human performance, machines already exceed human capabilities in areas where human cognitive limit our performance. For example, machines could theoretically handle hundreds of languages and process vast amounts of information on the Internet. Can future Humanoid Robotic assistants (in addition to handling mechanical chores) perhaps provide language and cultural support in a seamless and natural fashion?

    In this talk, I will present results from our efforts to make this dream become reality. We will discuss and demonstrate various multi- and cross-lingual communication and interpretation devices and services that we have deployed into use already, including multilingual dialog systems, pocket speech translators on smartphones, and automatic simultaneous interpretation services. I will then discuss the many open issues remaining from basic performance issues in speech and language processing, to providing better and more natural forms of deployment.

    Depending on scenario and interaction style, rather different strategies appear to be effective, requiring different degrees of sophistication of the user interfaces, speech, translation and dialog processing.

  • From Embodied Intelligence To Fetal Development – A Quest for The Fundamentals of Human-oid Intelligence Authors: Yasuo Kuniyoshi, The University of Tokyo, Japan
    Despite the vast advancement of methodologies and computational power, robot intelligence is still far from human level, leaving “hard problems” such as meaning and consciousness unsolved. Understanding the fundamentals of human intelligence is becoming more and more important. In this talk, I will first show that physics of human-like body in action already provide certain information structure which can set the natural basis of categorization and meaning. Then, I will show a principle of autonomous exploration that reveals the embodied information structure, aka. body affordances. Coordinated motor patterns consistent with the embodiment emerge from multiple chaotic elements coupled through body-environment physics.

    In humans, the above principle drives early motor development. And the resulting sensory-motor information can be captured by self-organizing neural circuits, forming the basis of cognitive structure. In order to investigate this hypothetical scenario, we constructed a simulation model of a human-oid fetus. It consists of a musculo-skeletal body, the environment within uterus, a neural model of spine and medulla, and a self-organizing neural network modelling sensory and motor cortical areas. A series of experiments revealed that with very little “innate” neural circuits, the model acquires various behaviour patterns that comply with its embodiment, and the neural model self organizes to capture the information structure.

    In 2012, a new five-year project called “Constructive Developmental Science” has started. It is a collaborative effort by roboticists, psychologists, medical scientists, and “Tojisha”, the researchers who investigate their own developmental disorders as internal observers. The fetal development simulation plays a core role there as a platform for integrating the interdisciplinary data and experimenting on the effect of various early genetic/environmental conditions on the courses of development. The project aims at revealing the fundamental principles of human development and new understanding of developmental disorders.

  • Rethinking Industrial Robots Authors: Rodney Brooks, Professor (emiritus) MIT and Founder, Chairman, and CTO of Rethink Robotics, Inc.
    At Rethink Robotics we have spent the last few years developing a new class of industrial robot.  Rather than being evaluated on speed, precision or repeatability, our robot, Baxter, is about flexibility, adaptability and ease of use.  It has built in safety and common sense behaviors, and is very inexpensive.  An ordinary factory worker can learn in a few minutes how to train it for new tasks.  The robot is aware of its surroundings and understands certain classes of objects in the world. The robot is the interface and it is trained at a task level through a person moving its arms and showing it objects.  It is a collaborative robot and can easily adapt to people changing the world around it, or even stepping in and doing part of its task themselves.  The Baxter robot is now in mass production and being sold to end-user customers.
  • Teaching Robots to Cook, Relax and Play Catch Authors: Aude Billard, EPFL
    "Teaching robots to do tasks is a more challenging task than it may appear at a first glance, even when these robots resemble us. In this presentation, I will shed some light on some of the major challenges we face when transferring skills from humans to robots.
    Imitation learning is not just play and record. A key component of learning is the ability to generalize. In imitation learning, generalization must come from very few examples to ensure that the length of the training phase is bearable to the human teacher. Most approaches, however, discard erroneous training examples, hence forcing the human teacher to be very skilled at the task, and to repeat the tasks numerous times until enough successful demonstrations have been gathered. There is an advantage to include errors in the training to ensure robustness and better generalization. I will hence discuss some promising avenues in learning from failed demonstrations.

    Traditional planning approaches seek to find the optimal solution. The strength of human control, however, lies in that we are capable of performing the same tasks in multiple ways, several of which being sub-optimal. Feasibility is hence more important than optimality. It offers the possibility to rapidly switch across control strategies in the face of perturbations. Modeling the variability with which humans perform the same task provides robots with a notion of feasibility regions in sensori-motor space. Over the years, we have sought to provide robots with controllers that allow instantaneous reactions to perturbation, mimicking humans’ immediate response in the presence of danger. I will describe application of such fast and robust adaptation for compliant control during manipulation of fragile objects and for performing sports, such as when playing golf with moving targets and when catching fast flying objects.

    "