View Chapter

Chapter 74 — Learning from Humans

Aude G. Billard, Sylvain Calinon and Rüdiger Dillmann

This chapter surveys the main approaches developed to date to endow robots with the ability to learn from human guidance. The field is best known as robot programming by demonstration, robot learning from/by demonstration, apprenticeship learning and imitation learning. We start with a brief historical overview of the field. We then summarize the various approaches taken to solve four main questions: when, what, who and when to imitate. We emphasize the importance of choosing well the interface and the channels used to convey the demonstrations, with an eye on interfaces providing force control and force feedback. We then review algorithmic approaches to model skills individually and as a compound and algorithms that combine learning from human guidance with reinforcement learning. We close with a look on the use of language to guide teaching and a list of open issues.

Learning from failure II

Author  Aude Billard

Video ID : 477

This video illustrates in a second example how learning from demonstration can benefit from failed demonstrations (as opposed to learning from successful demonstrations). Here, the robot Robota must learn how to coordinate its two arms in a timely manner for the left arm to hit the ball with the racket right on time, after the left arm sent the ball flying by hitting the catapult. More details on this work is available in: A. Rai, G. de Chambrier, A. Billard: Learning from failed demonstrations in unreliable systems, Proc. IEEE-RAS Int. Conf. Humanoid Robots (Humanoids), Atlanta (2013), pp. 410 – 416; doi: 10.1109/HUMANOIDS.2013.7030007 .

Chapter 19 — Robot Hands

Claudio Melchiorri and Makoto Kaneko

Multifingered robot hands have a potential capability for achieving dexterous manipulation of objects by using rolling and sliding motions. This chapter addresses design, actuation, sensing and control of multifingered robot hands. From the design viewpoint, they have a strong constraint in actuator implementation due to the space limitation in each joint. After briefly introducing the overview of anthropomorphic end-effector and its dexterity in Sect. 19.1, various approaches for actuation are provided with their advantages and disadvantages in Sect. 19.2. The key classification is (1) remote actuation or build-in actuation and (2) the relationship between the number of joints and the number of actuator. In Sect. 19.3, actuators and sensors used for multifingered hands are described. In Sect. 19.4, modeling and control are introduced by considering both dynamic effects and friction. Applications and trends are given in Sect. 19.5. Finally, this chapter is closed with conclusions and further reading.

The DLR Hand performing several tasks

Author  DLR - Robotics and Mechatronics Center

Video ID : 769

In the video, several experiments and the execution of different tasks by the DLR Hand II are shown.

Chapter 21 — Actuators for Soft Robotics

Alin Albu-Schäffer and Antonio Bicchi

Although we do not know as yet how robots of the future will look like exactly, most of us are sure that they will not resemble the heavy, bulky, rigid machines dangerously moving around in old fashioned industrial automation. There is a growing consensus, in the research community as well as in expectations from the public, that robots of the next generation will be physically compliant and adaptable machines, closely interacting with humans and moving safely, smoothly and efficiently - in other terms, robots will be soft.

This chapter discusses the design, modeling and control of actuators for the new generation of soft robots, which can replace conventional actuators in applications where rigidity is not the first and foremost concern in performance. The chapter focuses on the technology, modeling, and control of lumped parameters of soft robotics, that is, systems of discrete, interconnected, and compliant elements. Distributed parameters, snakelike and continuum soft robotics, are presented in Chap. 20, while Chap. 23 discusses in detail the biomimetic motivations that are often behind soft robotics.

Variable impedance actuators: Moving the robots of tomorrow

Author  B. Vanderborght, A. Albu-Schäffer, A. Bicchi, E. Burdet, D. Caldwell, R. Carloni, M. Catalano, Ganesh, Garabini, Grebenstein, Grioli, Haddadin, Jafari, Laffranchi, Lefeber, Petit, Stramigioli, Tsagarakis, Van Damme, Van Ham, Visser, Wolf

Video ID : 456

Most of today's robots have rigid structures and actuators requiring complex software control algorithms and sophisticated sensor systems in order to behave in a compliant and safe way adapted to contact with unknown environments and humans. By studying and constructing variable impedance actuators and their control, we contribute to the development of actuation units that can match the intrinsic safety, motion performance and energy efficiency of biological systems and, in particular, of the humans. As such, this may lead to a new generation of robots that can co-exist and co-operate with people and get closer to the human manipulation and locomotion performance than is possible with current robots.

Chapter 41 — Active Manipulation for Perception

Anna Petrovskaya and Kaijen Hsiao

This chapter covers perceptual methods in which manipulation is an integral part of perception. These methods face special challenges due to data sparsity and high costs of sensing actions. However, they can also succeed where other perceptual methods fail, for example, in poor-visibility conditions or for learning the physical properties of a scene.

The chapter focuses on specialized methods that have been developed for object localization, inference, planning, recognition, and modeling in activemanipulation approaches.We concludewith a discussion of real-life applications and directions for future research.

6-DOF object localization via touch

Author  Anna Petrovskaya

Video ID : 721

The PUMA robot arm performs 6-DOF localization of an object (i.e., a cash register) via touch starting with global uncertainty. After each contact, the robot analyzes the resulting belief about the object pose. If the uncertainty of the belief is too large, the robot continues to probe the object. Once, the uncertainty is small enough, the robot is able to push buttons and manipulate the drawer based on its knowledge of the object pose and prior knowledge of the object model. A prior 3-D mesh model of the object was constructed by touching the object with the robot's end-effector.

Chapter 23 — Biomimetic Robots

Kyu-Jin Cho and Robert Wood

Biomimetic robot designs attempt to translate biological principles into engineered systems, replacing more classical engineering solutions in order to achieve a function observed in the natural system. This chapter will focus on mechanism design for bio-inspired robots that replicate key principles from nature with novel engineering solutions. The challenges of biomimetic design include developing a deep understanding of the relevant natural system and translating this understanding into engineering design rules. This often entails the development of novel fabrication and actuation to realize the biomimetic design.

This chapter consists of four sections. In Sect. 23.1, we will define what biomimetic design entails, and contrast biomimetic robots with bio-inspired robots. In Sect. 23.2, we will discuss the fundamental components for developing a biomimetic robot. In Sect. 23.3, we will review detailed biomimetic designs that have been developed for canonical robot locomotion behaviors including flapping-wing flight, jumping, crawling, wall climbing, and swimming. In Sect. 23.4, we will discuss the enabling technologies for these biomimetic designs including material and fabrication.

RoACH: a 2.4 gram, untethered, crawling hexapod robot

Author  Aaron M. Hoover, Erik Steltz, Ronald S. Fearing

Video ID : 286

The robotic autonomous crawling hexapod (RoACH) is made using lightweight composites with integrated flexural hinges. It is actuated by two shape-memory-alloy wires and controlled by a PIC microprocessor. It can communicate over IrDA and run untethered for more than nine minutes on a single charge.

Chapter 34 — Visual Servoing

François Chaumette, Seth Hutchinson and Peter Corke

This chapter introduces visual servo control, using computer vision data in the servo loop to control the motion of a robot. We first describe the basic techniques that are by now well established in the field. We give a general overview of the formulation of the visual servo control problem, and describe the two archetypal visual servo control schemes: image-based and pose-based visual servo control. We then discuss performance and stability issues that pertain to these two schemes, motivating advanced techniques. Of the many advanced techniques that have been developed, we discuss 2.5-D, hybrid, partitioned, and switched approaches. Having covered a variety of control schemes, we deal with target tracking and controlling motion directly in the joint space and extensions to under-actuated ground and aerial robots. We conclude by describing applications of visual servoing in robotics.

PBVS on a 6-DOF robot arm (1)

Author  Francois Chaumette, Seth Hutchinson, Peter Corke

Video ID : 62

This video shows a PBVS on a 6-DOF robot arm with (c^t_o, theta u) as visual features. It corresponds to the results depicted in Figure 34.9.

Chapter 51 — Modeling and Control of Underwater Robots

Gianluca Antonelli, Thor I. Fossen and Dana R. Yoerger

This chapter deals with modeling and control of underwater robots. First, a brief introduction showing the constantly expanding role of marine robotics in oceanic engineering is given; this section also contains some historical backgrounds. Most of the following sections strongly overlap with the corresponding chapters presented in this handbook; hence, to avoid useless repetitions, only those aspects peculiar to the underwater environment are discussed, assuming that the reader is already familiar with concepts such as fault detection systems when discussing the corresponding underwater implementation. Themodeling section is presented by focusing on a coefficient-based approach capturing the most relevant underwater dynamic effects. Two sections dealing with the description of the sensor and the actuating systems are then given. Autonomous underwater vehicles require the implementation of mission control system as well as guidance and control algorithms. Underwater localization is also discussed. Underwater manipulation is then briefly approached. Fault detection and fault tolerance, together with the coordination control of multiple underwater vehicles, conclude the theoretical part of the chapter. Two final sections, reporting some successful applications and discussing future perspectives, conclude the chapter. The reader is referred to Chap. 25 for the design issues.

Mariana Trench: HROV Nereus samples the Challenger Deep seafloor

Author  Woods Hole Oceanographic Institution

Video ID : 89

Date: May 31, 2009. Depth: 10,006 meters (6.2 miles). A WHOI-led team successfully brought the newly-built hybrid remotely operated vehicle (HROV) Nereus to the deepest part of the world's ocean, the Challenger Deep in the Pacific Ocean. The dive makes the unmanned Nereus the world's deepest-diving vehicle and the first vehicle to explore the Mariana Trench since 1998. To learn more visit http://www.whoi.edu/page.do?pid=33775.

Chapter 11 — Robots with Flexible Elements

Alessandro De Luca and Wayne J. Book

Design issues, dynamic modeling, trajectory planning, and feedback control problems are presented for robot manipulators having components with mechanical flexibility, either concentrated at the joints or distributed along the links. The chapter is divided accordingly into two main parts. Similarities or differences between the two types of flexibility are pointed out wherever appropriate.

For robots with flexible joints, the dynamic model is derived in detail by following a Lagrangian approach and possible simplified versions are discussed. The problem of computing the nominal torques that produce a desired robot motion is then solved. Regulation and trajectory tracking tasks are addressed by means of linear and nonlinear feedback control designs.

For robots with flexible links, relevant factors that lead to the consideration of distributed flexibility are analyzed. Dynamic models are presented, based on the treatment of flexibility through lumped elements, transfer matrices, or assumed modes. Several specific issues are then highlighted, including the selection of sensors, the model order used for control design, and the generation of effective commands that reduce or eliminate residual vibrations in rest-to-rest maneuvers. Feedback control alternatives are finally discussed.

In each of the two parts of this chapter, a section is devoted to the illustration of the original references and to further readings on the subject.

Rest-to-rest motion for a flexible link

Author  Alessandro De Luca

Video ID : 779

This 2003 video shows a planar one-link flexible arm executing a desired rest-to-rest motion in a given finite time (90 deg in 2 s). Link deformations vanish completely at the desired final time. The applied control law is the combination of a model-based feedforward command designed for a smooth trajectory assigned to the flat output of the system and of a stabilizing PID feedback action on the joint angle around its associated trajectory. References: 1. A. De Luca, G. Di Giovanni: Rest-to-rest motion of a one-link flexible arm, Proc. IEEE/ASME Int. Conf. Adv. Intell. Mechatron., Como (2001), pp. 923-928; doi: 10.1109/AIM.2001.936793; 2. A. De Luca, V. Caiano, D. Del Vescovo: Experiments on rest-to-rest motion of a flexible arm, in B. Siciliano, P. Dario (Eds), Experimental Robotics VIII, Springer Tract. Adv. Robot. 5, 338-349 (2003); doi: 10.1007/3-540-36268-1_30

Chapter 76 — Evolutionary Robotics

Stefano Nolfi, Josh Bongard, Phil Husbands and Dario Floreano

Evolutionary Robotics is a method for automatically generating artificial brains and morphologies of autonomous robots. This approach is useful both for investigating the design space of robotic applications and for testing scientific hypotheses of biological mechanisms and processes. In this chapter we provide an overview of methods and results of Evolutionary Robotics with robots of different shapes, dimensions, and operation features. We consider both simulated and physical robots with special consideration to the transfer between the two worlds.

Evolved homing walk on rough ground

Author  Phil Husbands

Video ID : 373

Evolved, simulated hexapod walks over rough terrain while homing on a beacon. This behavior was incrementally evolved with the controlling neural-network architecture which was expanding at each stage. Work done at Sussex University by Eric Vaughan.

Chapter 71 — Cognitive Human-Robot Interaction

Bilge Mutlu, Nicholas Roy and Selma Šabanović

A key research challenge in robotics is to design robotic systems with the cognitive capabilities necessary to support human–robot interaction. These systems will need to have appropriate representations of the world; the task at hand; the capabilities, expectations, and actions of their human counterparts; and how their own actions might affect the world, their task, and their human partners. Cognitive human–robot interaction is a research area that considers human(s), robot(s), and their joint actions as a cognitive system and seeks to create models, algorithms, and design guidelines to enable the design of such systems. Core research activities in this area include the development of representations and actions that allow robots to participate in joint activities with people; a deeper understanding of human expectations and cognitive responses to robot actions; and, models of joint activity for human–robot interaction. This chapter surveys these research activities by drawing on research questions and advances from a wide range of fields including computer science, cognitive science, linguistics, and robotics.

Gaze and gesture cues for robots

Author  Bilge Mutlu

Video ID : 128

In human-robot communication, nonverbal cues like gaze and gesture can be a source of important information for starting and maintaining interaction. Gaze, for example, can tell a person about what the robot is attending to, its mental state, and its role in a conversation. Researchers are studying and developing models of nonverbal cues in human-robot interaction to enable more successful collaboration between robots and humans in a variety of domains, including education.