上海交通大学：《生物与机器人学》课程教学资源_Humanoid

1307 Humanc us 56,Humanoids Charles C.Kemp,Paul Fitzpatrick,Hirohisa Hirukawa,Kazuhito Yokoi,Kensuke Harada, Yoshio Matsumoto Humanoid robots selectively emulate aspects of 56.2 History and overview...........................1310 human form and behavior.Humanoids come 56.2.1 Different Forms...... 1311 56.2.2 Different Degrees of Freedom ...1311 in a variety of shapes and sizes,from complete 56.2.3 Different Sensors. 1311 human-size legged robots to isolated robotic heads 56.2.4 Other Dimensions of Variation......1312 with human-like sensing and expression.This chapter highlights significant humanoid platforms 56.3L0c0m0ti0n…….1312 56.3.1 Bipedal Locomotion. 1312 and achievements,and discusses some of the 56.3.2 Falling Down............... .1313 underlying goals behind this area of robotics. 56.3.3 Sensing for Balance............ .1314 Humanoids tend to require the integration of 56.3.4 Localization and Obstacle many of the methods covered in detail within Detection.... ..1314 other chapters of this handbook,so this chapter focuses on distinctive aspects of humanoid robotics 56.4 Manipulation..................... 1315 56.4.1 The Arm and Hand 1315 with liberal cross-referencing. 56.4.2 Sensing for Manipulation 1316 This chapter examines what motivates re- 56.4.3 Rhythmic Manipulation ..............1317 searchers to pursue humanoid robotics,and 56.4.4 Cooperative Manipulation 1317 provides a taste of the evolution of this field over 56.4.5 Learning and Development .1318 time.It summarizes work on legged humanoid 56.5 Whole-Body Activities....... .1318 locomotion,humanoid manipulation,whole-body 56.5.1 Coarse Whole-Body Motion 1319 activities,and approaches to human-robot com- 56.5.2 Generating Dynamically Stable munication.It concludes with a brief discussion of Motions..1321 factors that may influence the future of humanoid 56.5.3 Generating Whole-Body Motions robots. from Operational Point Motions....1322 56.5.4 Generating Motions 56.1 Why Humanoids?................. ..1307 when in Contact with an Object....1324 56.1.1 The Human Example ,1308 56.6 Communication.... ...1325 56.1.2 The Pleasing Mirror. .1308 56.6.1 Expressive Morphology 56.1.3 Understanding Intelligence..........1308 and Behavior.................1325 56.1.4 Interfacing 56.6.2 Interpreting Human Expression.....1327 with the Human World .1308 56.6.3 Alternative Models 56.1.5 Interfacing with People...............1309 for Human-Robot Communication 1329 56.1.6 Entertainment,Culture, 56.7 Conclusions and Further Reading....1329 and Surrogates 1310 References ........... 1329 56.1 Why Humanoids? Throughout history,the human body and mind have Chap.60,Biologically Inspired Robots).These robots inspired artists,engineers,and scientists.The field of usually share similar kinematics to humans,as well as humanoid robotics focuses on the creation of robots similar sensing and behavior.The motivations that have that are directly inspired by human capabilities (see driven the development of humanoid robots vary widely

1307 Humanoids 56. Humanoids Charles C. Kemp, Paul Fitzpatrick, Hirohisa Hirukawa, Kazuhito Yokoi, Kensuke Harada, Yoshio Matsumoto Humanoid robots selectively emulate aspects of human form and behavior. Humanoids come in a variety of shapes and sizes, from complete human-size legged robots to isolated robotic heads with human-like sensing and expression. This chapter highlights significant humanoid platforms and achievements, and discusses some of the underlying goals behind this area of robotics. Humanoids tend to require the integration of many of the methods covered in detail within other chapters of this handbook, so this chapter focuses on distinctive aspects of humanoid robotics with liberal cross-referencing. This chapter examines what motivates re￾searchers to pursue humanoid robotics, and provides a taste of the evolution of this field over time. It summarizes work on legged humanoid locomotion, humanoid manipulation, whole-body activities, and approaches to human–robot com￾munication. It concludes with a brief discussion of factors that may influence the future of humanoid robots. 56.1 Why Humanoids? .................................. 1307 56.1.1 The Human Example.................... 1308 56.1.2 The Pleasing Mirror...................... 1308 56.1.3 Understanding Intelligence .......... 1308 56.1.4 Interfacing with the Human World ................ 1308 56.1.5 Interfacing with People ............... 1309 56.1.6 Entertainment, Culture, and Surrogates ........................... 1310 56.2 History and Overview ............................ 1310 56.2.1 Different Forms ........................... 1311 56.2.2 Different Degrees of Freedom ....... 1311 56.2.3 Different Sensors ......................... 1311 56.2.4 Other Dimensions of Variation ...... 1312 56.3 Locomotion.......................................... 1312 56.3.1 Bipedal Locomotion..................... 1312 56.3.2 Falling Down .............................. 1313 56.3.3 Sensing for Balance ..................... 1314 56.3.4 Localization and Obstacle Detection ................................... 1314 56.4 Manipulation ....................................... 1315 56.4.1 The Arm and Hand ...................... 1315 56.4.2 Sensing for Manipulation ............. 1316 56.4.3 Rhythmic Manipulation ............... 1317 56.4.4 Cooperative Manipulation ............ 1317 56.4.5 Learning and Development .......... 1318 56.5 Whole-Body Activities ........................... 1318 56.5.1 Coarse Whole-Body Motion .......... 1319 56.5.2 Generating Dynamically Stable Motions ..................................... 1321 56.5.3 Generating Whole-Body Motions from Operational Point Motions .... 1322 56.5.4 Generating Motions when in Contact with an Object .... 1324 56.6 Communication .................................... 1325 56.6.1 Expressive Morphology and Behavior.............................. 1325 56.6.2 Interpreting Human Expression..... 1327 56.6.3 Alternative Models for Human–Robot Communication 1329 56.7 Conclusions and Further Reading ........... 1329 References .................................................. 1329 56.1 Why Humanoids? Throughout history, the human body and mind have inspired artists, engineers, and scientists. The field of humanoid robotics focuses on the creation of robots that are directly inspired by human capabilities (see Chap. 60, Biologically Inspired Robots). These robots usually share similar kinematics to humans, as well as similar sensing and behavior. The motivations that have driven the development of humanoid robots vary widely. Part G 56

1308 Part G Human-Centered and Life-Like Robotics ence fiction play Rossum's Universal Robots(R.U.R.) Part For example,humanoid robots have been developed as general-purpose mechanical workers,as entertain- is centered around the story of artificial people created G156. ers,and as test-beds for theories from neuroscience and in a factory [56.9].This play from 1920 is widely be- experimental psychology [56.1-3]. lieved to have popularized the term robot.Many other works have included explicit representations of hu- 56.1.1 The Human Example manoid robots,such as the robot Maria in Fritz Lang's 1927 film Metropolis [56.10],and the thoughtful por- On a daily basis,humans perform important tasks that trayal of humanoid robotics by Isaac Asimov in works are well beyond the capabilities of current robots.More- such as The Caves of Steel from 1954 [56.11].The long over,humans are generalists with the ability to perform history of humanoid robots in science fiction has influ- a wide variety of distinct tasks.Roboticists would like enced generations of researchers,as well as the general to create robots with comparable versatility and skill. public,and serves as further evidence that people are When automating a task that people perform,it is natu- drawn to the idea of humanoid robots. ral to consider the physical and intellectual mechanisms that enable a person to perform the task.Exactly what to 56.1.3 Understanding Intelligence borrow from the human example is controversial.The Many researchers in the humanoid robotics commu- literal-minded approach of creating humanoid robots nity see humanoid robots as a tool with which to better may not be the best way to achieve some human-like capabilities(see Chap.54,Domestic Robots).For exam- understand humans [56.3,12].Humanoid robots offer an ple,dishwashing machines bear little similarity to the avenue to test understanding through construction(syn- manual dishwashing they replace. thesis),and thereby complement the careful analysis provided by researchers in disciplines such as cognitive 56.1.2 The Pleasing Mirror science. Researchers have sought to better emulate human in- Humans are humanity's favorite subject.A quick look at telligence using humanoid robotics [56.13].Scientists, popular magazines,videos,and books should be enough developmental psychologists,and linguists have found to convince any alien observer that humanity is obsessed strong links between the human body and human cog- with itself.The nature of this obsession is not fully un- nition [56.14].By being embodied in a manner similar to humans,and situated within human environments, derstood,but aspects of it have influenced the field of humanoid robots may be able to exploit similar mech- humanoid robotics. anisms for artificial intelligence (AI).Researchers are Humans are social animals that generally like to ob- serve and interact with one another [56.4].Moreover, also attempting to find methods that will enable robots people are highly attuned to human characteristics,such to develop autonomously in a manner akin to human in- as the sound of human voices and the appearance of fants [56.15].Some of these researchers use humanoid human faces and body motion [56.5-7].Infants show robots that can physically explore the world in a manner preferences for these types of stimuli at a very young age, similar to humans [56.16]. and adults appear to use specialized mental resources 56.1.4 Interfacing with the Human World when interpreting these stimuli.By mimicking human characteristics,humanoid robots can engage these same Environments built for humans have been designed to ac- preferences and mental resources. commodate human form and behavior [56.17,18].Many Humanity's narcissism has been reflected in media important everyday objects fit in a person's hand and are as diverse as cave paintings,sculpture,mechanical toys,light enough to be transported conveniently by a per- photographs,and computer animation.Artists have con-son.Human tools match human dexterity.Doors tend sistently attempted to portray people with the latest tools to be a convenient size for people to walk through.Ta- at their disposal.Robotics serves as a powerful new bles and desks are at a height that is well matched to medium that enables the creation of artifacts that oper- the human body and senses.Humanoid robots can po- ate within the real world and exhibit both human form tentially take advantage of these same accommodations, and behavior [56.8]. thereby simplifying tasks and avoiding the need to al- Popular works of fiction have frequently included ter the environment for the robot [56.19].For example, influential portrayals of humanoid robots and manmade humanoid robots and people could potentially collabo- humanoid creatures.For example,Karel Capek's sci- rate with one another in the same space using the same

1308 Part G Human-Centered and Life-Like Robotics For example, humanoid robots have been developed as general-purpose mechanical workers, as entertain￾ers, and as test-beds for theories from neuroscience and experimental psychology [56.1–3]. 56.1.1 The Human Example On a daily basis, humans perform important tasks that are well beyond the capabilities of current robots. More￾over, humans are generalists with the ability to perform a wide variety of distinct tasks. Roboticists would like to create robots with comparable versatility and skill. When automating a task that people perform, it is natu￾ral to consider the physical and intellectual mechanisms that enable a person to perform the task. Exactly what to borrow from the human example is controversial. The literal-minded approach of creating humanoid robots may not be the best way to achieve some human-like capabilities (see Chap. 54, Domestic Robots). For exam￾ple, dishwashing machines bear little similarity to the manual dishwashing they replace. 56.1.2 The Pleasing Mirror Humans are humanity’s favorite subject. A quick look at popular magazines, videos, and books should be enough to convince any alien observer that humanity is obsessed with itself. The nature of this obsession is not fully un￾derstood, but aspects of it have influenced the field of humanoid robotics. Humans are social animals that generally like to ob￾serve and interact with one another [56.4]. Moreover, people are highly attuned to human characteristics, such as the sound of human voices and the appearance of human faces and body motion [56.5–7]. Infants show preferences for these types of stimuli at a very young age, and adults appear to use specialized mental resources when interpreting these stimuli. By mimicking human characteristics, humanoid robots can engage these same preferences and mental resources. Humanity’s narcissism has been reflected in media as diverse as cave paintings, sculpture, mechanical toys, photographs, and computer animation. Artists have con￾sistently attempted to portray people with the latest tools at their disposal. Robotics serves as a powerful new medium that enables the creation of artifacts that oper￾ate within the real world and exhibit both human form and behavior [56.8]. Popular works of fiction have frequently included influential portrayals of humanoid robots and manmade humanoid creatures. For example, Karel Capek’s sci- ˘ ence fiction play Rossum’s Universal Robots (R.U.R.) is centered around the story of artificial people created in a factory [56.9]. This play from 1920 is widely be￾lieved to have popularized the term robot. Many other works have included explicit representations of hu￾manoid robots, such as the robot Maria in Fritz Lang’s 1927 film Metropolis [56.10], and the thoughtful por￾trayal of humanoid robotics by Isaac Asimov in works such as The Caves of Steel from 1954 [56.11]. The long history of humanoid robots in science fiction has influ￾enced generations of researchers, as well as the general public, and serves as further evidence that people are drawn to the idea of humanoid robots. 56.1.3 Understanding Intelligence Many researchers in the humanoid robotics commu￾nity see humanoid robots as a tool with which to better understand humans [56.3,12]. Humanoid robots offer an avenue to test understanding through construction (syn￾thesis), and thereby complement the careful analysis provided by researchers in disciplines such as cognitive science. Researchers have sought to better emulate human in￾telligence using humanoid robotics [56.13]. Scientists, developmental psychologists, and linguists have found strong links between the human body and human cog￾nition [56.14]. By being embodied in a manner similar to humans, and situated within human environments, humanoid robots may be able to exploit similar mech￾anisms for artificial intelligence (AI). Researchers are also attempting to find methods that will enable robots to develop autonomously in a manner akin to human in￾fants [56.15]. Some of these researchers use humanoid robots that can physically explore the world in a manner similar to humans [56.16]. 56.1.4 Interfacing with the Human World Environments built for humans have been designed to ac￾commodate human form and behavior [56.17,18]. Many important everyday objects fit in a person’s hand and are light enough to be transported conveniently by a per￾son. Human tools match human dexterity. Doors tend to be a convenient size for people to walk through. Ta￾bles and desks are at a height that is well matched to the human body and senses. Humanoid robots can po￾tentially take advantage of these same accommodations, thereby simplifying tasks and avoiding the need to al￾ter the environment for the robot [56.19]. For example, humanoid robots and people could potentially collabo￾rate with one another in the same space using the same Part G 56.1

Humanoids 56.1 Why Humanoids? 1309 Part G56. Fig.56.3 The humanoid robot HRP-2 dancing with a human [56.22]. The human is a master of a traditional Japanese dance whose danc- Fig.56.1 The humanoid robot HRP-IS driving a backhoe ing was recorded by a motion-capture system,and transformed for (Courtesy of Kawasaki Heavy Industries,Tokyu Construc- use by the robot tion and AIST).The robot can be teleoperated by a human operator to control the backhoe remotely.The same robot change its posture in order to lean into something, could potentially interface with many different unmodified pull with the weight of its body,or crawl under an machines obstacle [56.24,251. tools [56.20].Humanoid robots can also interface with 56.1.5 Interfacing with People machinery that does not include drive-by-wire controls, as shown by the teleoperated robot in the cockpit of People are accustomed to working with other people. a backhoe in Fig.56.1 [56.211. Many types of communication rely on human form and Mobility serves as another example.It is very diffi-behavior.Some types of natural gestures and expres- cult to create a tall wheeled robot with a small footprint sion involve subtle movements in the hands and face that is capable of traversing stairs and moving over (Chap.58,Social Robots that Interact with People).Peo- rough terrain.Robots with legs and human-like behav- ple can interpret eye gaze and facial expressions without ior could potentially traverse the same environments training.Humanoid robots can potentially simplify and that humans traverse,such as the industrial plant shown enhance human-robot interaction by taking advantage of in Fig.56.2,which has stairs and handrails designed the communications channels that already exist between for human use [56.23].In addition to mobility advan-people. tages,legs have the potential to help in other ways. Similarly,people already have the ability to per- For example,legs could enable a humanoid robot to form many desirable tasks.This task knowledge may Fig.56.2 HRP-1 operating in a mockup of an industrial Fig.56.4 Actroid (Courtesy of Kokoro),an android de- plant(Courtesy of Mitsubishi Heavy Industries) signed for entertainment,telepresence,and media roles

Humanoids 56.1 Why Humanoids? 1309 Fig. 56.1 The humanoid robot HRP-1S driving a backhoe (Courtesy of Kawasaki Heavy Industries, Tokyu Construc￾tion and AIST). The robot can be teleoperated by a human operator to control the backhoe remotely. The same robot could potentially interface with many different unmodified machines tools [56.20]. Humanoid robots can also interface with machinery that does not include drive-by-wire controls, as shown by the teleoperated robot in the cockpit of a backhoe in Fig. 56.1 [56.21]. Mobility serves as another example. It is very diffi- cult to create a tall wheeled robot with a small footprint that is capable of traversing stairs and moving over rough terrain. Robots with legs and human-like behav￾ior could potentially traverse the same environments that humans traverse, such as the industrial plant shown in Fig. 56.2, which has stairs and handrails designed for human use [56.23]. In addition to mobility advan￾tages, legs have the potential to help in other ways. For example, legs could enable a humanoid robot to Fig. 56.2 HRP-1 operating in a mockup of an industrial plant (Courtesy of Mitsubishi Heavy Industries) Fig. 56.3 The humanoid robot HRP-2 dancing with a human [56.22]. The human is a master of a traditional Japanese dance whose danc￾ing was recorded by a motion-capture system, and transformed for use by the robot change its posture in order to lean into something, pull with the weight of its body, or crawl under an obstacle [56.24, 25]. 56.1.5 Interfacing with People People are accustomed to working with other people. Many types of communication rely on human form and behavior. Some types of natural gestures and expres￾sion involve subtle movements in the hands and face (Chap. 58, Social Robots that Interact with People). Peo￾ple can interpret eye gaze and facial expressions without training. Humanoid robots can potentially simplify and enhance human–robot interaction by taking advantage of the communications channels that already exist between people. Similarly, people already have the ability to per￾form many desirable tasks. This task knowledge may Fig. 56.4 Actroid (Courtesy of Kokoro), an android de￾signed for entertainment, telepresence, and media roles Part G 56.1

1310 Part G Human-Centered and Life-Like Robotics be more readily transferred to humanoid robots than to sembles a human,see Fig.56.4.For a humanoid robot Part G156 a robot with a drastically different body.This is espe- to be an improvement over a wax figure or an anima- cially true of cultural actions centered around the human tronic historical character,it must be realistic in form form (Fig.56.3). and function. People may one day wish to have robots that can 56.1.6 Entertainment,Culture, serve as an avatar for telepresence,model clothing,test and Surrogates ergonomics,or serve other surrogate roles that funda- mentally depend on the robot's similarity to a person. Humanoid robots are inherently appropriate for some Along these lines,robotic prosthetics have a close rela- applications.For example,many potential forms of en- tionship to humanoid robotics,since they seek to directly tertainment,such as theater,theme parks,and adult replace parts of the human body in form and function companionship,would rely on a robot that closely re- (Chap.53,Health Care and Rehabilitation Robotics). 56.2 History and Overview There is a long history of mechanical systems with and in Japan there is a tradition of creating mechanical human form that perform human-like movements.For dolls called Karakuri ningyo that dates back to at least example,Al-Jazari designed a humanoid automaton in the 18th century [56.29].In the 20th century,animatron- the 13th century [56.27],Leonardo da Vinci designed ics became an attraction at theme parks.For example, a humanoid automaton in the late 15th century [56.28], in 1967 Disneyland opened its Pirate's of the Caribbean ride [56.30],which featured animatronic pirates that play back human-like movements synchronized with audio. Although programmable,these humanoid animatronic systems moved in a fixed open-loop fashion without sensing their environment. In the second half of the 20th century,advances in digital computing enabled researchers to incorporate significant computation into their robots for sensing, control,and actuation.Many roboticists developed iso- lated systems for sensing.,locomotion,and manipulation that were inspired by human capabilities.However,the first humanoid robot to integrate all of these functions and capture widespread attention was WABOT-1,devel- oped by Ichiro Kato et al.at Waseda University in Japan Fig.56.5 WABOT-1(1973)and WABOT-2(1984)(Cour- in1973Fig.56.5). tesy of Humanoid Robotics Institute,Waseda University) The WABOT robots integrated functions that have been under constant elaboration since:visual object recognition,speech generation,speech recognition, bimanual object manipulation,and bipedal walking. WABOT-2's ability to play a piano,publicized at the Tsukuba Science Expo in 1985,stimulated significant public interest. In 1986,Honda began a confidential project to create a humanoid biped.Honda grew interested in humanoids, perhaps seeing in them devices of complexity compar- able to cars with the potential to become high-volume consumer products one day.In 1996,Honda unveiled Fig.56.6 Honda P2 (180 cm tall,210kg),P3 (160 cm,130kg),and the Honda Humanoid P2,the result of this confidential Asimo (120 cm,43 kg)[56.26].(Images courtesy of Honda) project.P2 was the first full-scale humanoid capable of

1310 Part G Human-Centered and Life-Like Robotics be more readily transferred to humanoid robots than to a robot with a drastically different body. This is espe￾cially true of cultural actions centered around the human form (Fig. 56.3). 56.1.6 Entertainment, Culture, and Surrogates Humanoid robots are inherently appropriate for some applications. For example, many potential forms of en￾tertainment, such as theater, theme parks, and adult companionship, would rely on a robot that closely re￾sembles a human, see Fig. 56.4. For a humanoid robot to be an improvement over a wax figure or an anima￾tronic historical character, it must be realistic in form and function. People may one day wish to have robots that can serve as an avatar for telepresence, model clothing, test ergonomics, or serve other surrogate roles that funda￾mentally depend on the robot’s similarity to a person. Along these lines, robotic prosthetics have a close rela￾tionship to humanoid robotics, since they seek to directly replace parts of the human body in form and function (Chap. 53, Health Care and Rehabilitation Robotics). 56.2 History and Overview There is a long history of mechanical systems with human form that perform human-like movements. For example, Al-Jazari designed a humanoid automaton in the 13th century [56.27], Leonardo da Vinci designed a humanoid automaton in the late 15th century [56.28], Fig. 56.5 WABOT-1 (1973) and WABOT-2 (1984) (Cour￾tesy of Humanoid Robotics Institute, Waseda University) Fig. 56.6 Honda P2 (180 cm tall, 210 kg), P3 (160 cm, 130 kg), and Asimo (120 cm, 43 kg) [56.26]. (Images courtesy of Honda) and in Japan there is a tradition of creating mechanical dolls called Karakuri ningyo that dates back to at least the 18th century [56.29]. In the 20th century, animatron￾ics became an attraction at theme parks. For example, in 1967 Disneyland opened its Pirate’s of the Caribbean ride [56.30], which featured animatronic pirates that play back human-like movements synchronized with audio. Although programmable, these humanoid animatronic systems moved in a fixed open-loop fashion without sensing their environment. In the second half of the 20th century, advances in digital computing enabled researchers to incorporate significant computation into their robots for sensing, control, and actuation. Many roboticists developed iso￾lated systems for sensing, locomotion, and manipulation that were inspired by human capabilities. However, the first humanoid robot to integrate all of these functions and capture widespread attention was WABOT-1, devel￾oped by Ichiro Kato et al. at Waseda University in Japan in 1973 (Fig. 56.5). The WABOT robots integrated functions that have been under constant elaboration since: visual object recognition, speech generation, speech recognition, bimanual object manipulation, and bipedal walking. WABOT-2’s ability to play a piano, publicized at the Tsukuba Science Expo in 1985, stimulated significant public interest. In 1986, Honda began a confidential project to create a humanoid biped. Honda grew interested in humanoids, perhaps seeing in them devices of complexity compar￾able to cars with the potential to become high-volume consumer products one day. In 1996, Honda unveiled the Honda Humanoid P2, the result of this confidential project. P2 was the first full-scale humanoid capable of Part G 56.2

Humanoids 56.2 History and Overview 1311 stable bipedal walking with onboard power and process- ing.Successive designs reduced its weight and improved performance (see Fig.56.6).Compared to humanoids built by academic laboratories and small manufacturers, the Honda humanoids were a leap forward in stur- Part G56.2 diness,using specially cast lightweight high-rigidity mechanical links,and harmonic drives with high torque capacity. In parallel with these developments,the decade- long Cog project began in 1993 at the MIT Artificial Intelligence laboratory in the USA with the intention of creating a humanoid robot that would,learn to think'by building on its bodily experiences to accom- plish progressively more abstract tasks [56.13].This project gave rise to an upper-body humanoid robot Fig.56.8 The NASA Robonaut consists of an upper body whose design was heavily inspired by the biological placed on a wheeled mobile base and cognitive sciences.Since the inception of the Cog project,many humanoid robotics projects with similar ety of humanoid robots that selectively emphasize some objectives have been initiated,and communities focused human characteristics,while deviating from others on developmental robotics,autonomous mental devel- One of the most noticeable axes of variation in hu- opment (AMD [56.31]),and epigenetic robotics have manoid robots is the presence or absence of body parts. emerged [56.32]. Some humanoid robots have focused solely on the head As of the early 21st century,many companies andand face,others have a head with two arms mounted to academic researchers have become involved with hu-a stationary torso,or a torso with wheels(see,for ex- manoid robots,and there are numerous humanoid robots ample,Fig.56.8),and still others have an articulate and across the world with distinctive features. expressive face with arms,legs,and a torso.Clearly,this variation in form impacts the ways in which the robot can 56.2.1 Different Forms be used,especially in terms of mobility,manipulation, whole-body activities.and human-robot interaction. Today,humanoid robots come in a variety of shapes and sizes that emulate different aspects of human form and 56.2.2 Different Degrees of Freedom behavior(Fig.56.7).As discussed,the motivations that have driven the development of humanoid robots vary Humanoid robots also tend to emulate some degrees widely.These diverse motivations have lead to a vari- of freedom in the human body,while ignoring others. Humanoid robots focusing on facial expressivity often incorporate actuated degrees of freedom in the face to generate facial expressions akin to those that humans can generate with their facial muscles.Likewise,the upper body of humanoid robots usually includes two arms,each with a one-degree-of-freedom (one-DOF) rotary joint at the elbow and a three-DOF rotary joint for the shoulder,but rarely attempt to emulate the human shoulder's ability to translate or the flexibility of the human spine [56.33,341. In general,humanoid robots tend to have a large number of degrees of freedom and a kinematic structure that may not be amenable to closed-form analysis due to redundancy and the lack of a closed-form inverse.This is in contrast to traditional industrial manipulators that Fig.56.7 Kismet is an example of a humanoid head for are often engineered to have minimal redundancy (six social interaction DOFs)and more easily analyzed kinematic structures

Humanoids 56.2 History and Overview 1311 stable bipedal walking with onboard power and process￾ing. Successive designs reduced its weight and improved performance (see Fig. 56.6). Compared to humanoids built by academic laboratories and small manufacturers, the Honda humanoids were a leap forward in stur￾diness, using specially cast lightweight high-rigidity mechanical links, and harmonic drives with high torque capacity. In parallel with these developments, the decade￾long Cog project began in 1993 at the MIT Artificial Intelligence laboratory in the USA with the intention of creating a humanoid robot that would, learn to ‘think’ by building on its bodily experiences to accom￾plish progressively more abstract tasks [56.13]. This project gave rise to an upper-body humanoid robot whose design was heavily inspired by the biological and cognitive sciences. Since the inception of the Cog project, many humanoid robotics projects with similar objectives have been initiated, and communities focused on developmental robotics, autonomous mental devel￾opment (AMD [56.31]), and epigenetic robotics have emerged [56.32]. As of the early 21st century, many companies and academic researchers have become involved with hu￾manoid robots, and there are numerous humanoid robots across the world with distinctive features. 56.2.1 Different Forms Today, humanoid robots come in a variety of shapes and sizes that emulate different aspects of human form and behavior (Fig. 56.7). As discussed, the motivations that have driven the development of humanoid robots vary widely. These diverse motivations have lead to a vari￾Fig. 56.7 Kismet is an example of a humanoid head for social interaction Fig. 56.8 The NASA Robonaut consists of an upper body placed on a wheeled mobile base ety of humanoid robots that selectively emphasize some human characteristics, while deviating from others. One of the most noticeable axes of variation in hu￾manoid robots is the presence or absence of body parts. Some humanoid robots have focused solely on the head and face, others have a head with two arms mounted to a stationary torso, or a torso with wheels (see, for ex￾ample, Fig. 56.8), and still others have an articulate and expressive face with arms, legs, and a torso. Clearly, this variation in form impacts the ways in which the robot can be used, especially in terms of mobility, manipulation, whole-body activities, and human–robot interaction. 56.2.2 Different Degrees of Freedom Humanoid robots also tend to emulate some degrees of freedom in the human body, while ignoring others. Humanoid robots focusing on facial expressivity often incorporate actuated degrees of freedom in the face to generate facial expressions akin to those that humans can generate with their facial muscles. Likewise, the upper body of humanoid robots usually includes two arms, each with a one-degree-of-freedom (one-DOF) rotary joint at the elbow and a three-DOF rotary joint for the shoulder, but rarely attempt to emulate the human shoulder’s ability to translate or the flexibility of the human spine [56.33, 34]. In general, humanoid robots tend to have a large number of degrees of freedom and a kinematic structure that may not be amenable to closed-form analysis due to redundancy and the lack of a closed-form inverse. This is in contrast to traditional industrial manipulators that are often engineered to have minimal redundancy (six DOFs) and more easily analyzed kinematic structures. Part G 56.2

1312 Part G Human-Centered and Life-Like Robotics 56.2.3 Different Sensors the robot will see a different world than the human.With Part G156 respect to behavior,placement of sensors on the head of Humanoid robots have made use of a variety of sen- the robot allows the robot to sense the world from a van- sors including cameras,laser range finders,microphone tage point that is similar to that of a human,which can arrays,lavalier microphones,and pressure sensors. w be important for finding objects that are sitting on a desk Some researchers choose to emulate human sensing by or table. selecting sensors with clear human analogs and mount- Prominent humanoid robots have added additional ing these sensors on the humanoid robot in a manner that sensors without human analogs.For example,Kismet mimics the placement of human sensory organs.As dis- used a camera mounted in its forehead to augment the cussed in Sect.56.6,this is perhaps most evident in the two cameras in its servoed eyes,which simplified com- use of cameras.Two to four cameras are often mounted mon tasks such as tracking faces.Similarly,versions of within the head of a humanoid robot with a configuration Asimo have used a camera mounted on its lower torso similar to human eyes. that looks down at the floor in order to simplify obstacle The justifications for this bias towards human-like detection and navigation during locomotion. sensing include the impact of sensing on natural human- robot interaction,the proven ability of the human senses 56.2.4 Other Dimensions of Variation to support human behavior,and aesthetics.For example, with respect to human-robot interaction,nonexperts can Other significant forms of variation include the size of sometimes interpret the functioning and implications of the robot.the method of actuation.the extent to which the a human-like sensor.such as a camera,more easily.Sim- robot attempts to appear like a human,and the activities ilarly,if a robot senses infrared or ultraviolet radiation, the robot performs. 56.3 Locomotion Bipedal walking is a key research topic in humanoid typically need to balance dynamically when walking robotics (see also Chap.16,Legged Robots,for a review bipedally. of this topic in the context of locomotion in general). Legged locomotion is a challenging area of robotics 56.3.1 Bipedal Locomotion research,and bipedal humanoid locomotion is espe- cially challenging.Some small humanoid robots are Currently the dominant methods for bipedal legged loco- able to achieve statically stable gaits by having large motion with humanoids make use of the zero-moment feet and a low center of mass,but large humanoids with point(ZMP)criterion to ensure that the robot does not a human-like weight distribution and body dimensions fall over [56.35].As discussed in detail in Chap.16,con- 1=0s t=2.5s 1=5s t=7.5s 1=17.5s t=15s t=12.5s t=10s Fig.56.9 HRP-2 walks on a slightly uneven surface

1312 Part G Human-Centered and Life-Like Robotics 56.2.3 Different Sensors Humanoid robots have made use of a variety of sen￾sors including cameras, laser range finders, microphone arrays, lavalier microphones, and pressure sensors. Some researchers choose to emulate human sensing by selecting sensors with clear human analogs and mount￾ing these sensors on the humanoid robot in a manner that mimics the placement of human sensory organs. As dis￾cussed in Sect. 56.6, this is perhaps most evident in the use of cameras. Two to four cameras are often mounted within the head of a humanoid robot with a configuration similar to human eyes. The justifications for this bias towards human-like sensing include the impact of sensing on natural human– robot interaction, the proven ability of the human senses to support human behavior, and aesthetics. For example, with respect to human–robot interaction, nonexperts can sometimes interpret the functioning and implications of a human-like sensor, such as a camera, more easily. Sim￾ilarly, if a robot senses infrared or ultraviolet radiation, the robot will see a different world than the human. With respect to behavior, placement of sensors on the head of the robot allows the robot to sense the world from a van￾tage point that is similar to that of a human, which can be important for finding objects that are sitting on a desk or table. Prominent humanoid robots have added additional sensors without human analogs. For example, Kismet used a camera mounted in its forehead to augment the two cameras in its servoed eyes, which simplified com￾mon tasks such as tracking faces. Similarly, versions of Asimo have used a camera mounted on its lower torso that looks down at the floor in order to simplify obstacle detection and navigation during locomotion. 56.2.4 Other Dimensions of Variation Other significant forms of variation include the size of the robot, the method of actuation, the extent to which the robot attempts to appear like a human, and the activities the robot performs. 56.3 Locomotion Bipedal walking is a key research topic in humanoid robotics (see also Chap. 16, Legged Robots, for a review of this topic in the context of locomotion in general). Legged locomotion is a challenging area of robotics research, and bipedal humanoid locomotion is espe￾cially challenging. Some small humanoid robots are able to achieve statically stable gaits by having large feet and a low center of mass, but large humanoids with a human-like weight distribution and body dimensions t=17.5 s t=0 s t=15 s t=2.5 s t=12.5 s t=5 s t=10 s t=7.5 s Fig. 56.9 HRP-2 walks on a slightly uneven surface typically need to balance dynamically when walking bipedally. 56.3.1 Bipedal Locomotion Currently the dominant methods for bipedal legged loco￾motion with humanoids make use of the zero-moment point (ZMP) criterion to ensure that the robot does not fall over [56.35]. As discussed in detail in Chap. 16, con￾Part G 56.3

Humanoids 56.3 Locomotion 1313 Part G56. w Fig.56.12 The humanoid robot HRP-2P getting up from a lying- Fig.56.10 These robots from Delft,MIT and Cornell (left down position to right)are designed to exploit their natural dynamics when walking [56.36].(Image courtesy of Steven H.Collins) to this task.So far,a generic and rigorous new criterion has not been established. trol of the robot's body such that the ZMP sits within the As an example of an alternative mode of bipedal support polygon of the robot's foot ensures that the foot locomotion,some running robots have used controllers remains planted on the ground,assuming that friction is based on an inverted pendulum model to achieve sta- high enough to avoid slipping.The ZMP can be used to ble gaits.These methods change the landing positions plan motion patterns that make the robot dynamically to keep the robot dynamically stable [56.37].More re- stable while walking. cently,researchers have begun to use the principles Controllers based on the ZMP criterion try to fol-of bipedal passive-dynamic walkers to develop pow- low a planned sequence of contact states and are often ered bipedal walkers that walk with high efficiency unable to change the landing positions in real time inin a human-like way by exploiting natural dynam- response to lost contact.Current ZMP-based bipedal ics(Fig.56.10 [56.361). walking algorithms have difficulty handling unexpected perturbations,such as might be encountered with un- 56.3.2 Falling Down even natural terrain (Fig.56.9).Robots using ZMP differ from human locomotion in significant ways.For exam- A human-scale robot should expect to fall from time ple,unlike people,robots using ZMP typically do not to time in realistic conditions.A humanoid robot may exploit the natural dynamics of their legs,or control the fall down due to a large disturbance even if the motion impedance of their joints. is planned carefully and a sophisticated feedback con- Augmenting ZMP-based control is currently an ac- troller is applied to the robot.In this event,the robot tive area of research.As will be discussed in Sect.56.5 could be damaged significantly during a fall,and could on whole-body activities researchers are working to also damage the environment or injure people who are integrate manipulation and bipedal locomotion.For ex-nearby.An important area of research is how to control ample,when a robot walks while grasping a handrail,the robot's fall in order to gracefully recover or minimize the contact could potentially increase the stability of the damage.The Sony QRIO can control its falling motions robot,but the ZMP criterion does not easily generalize in order to reduce the impact of touch down [56.38],al- Fig.56.11 Example of controlled falling-down motion

Humanoids 56.3 Locomotion 1313 Fig. 56.10 These robots from Delft, MIT and Cornell (left to right) are designed to exploit their natural dynamics when walking [56.36]. (Image courtesy of Steven H. Collins) trol of the robot’s body such that the ZMP sits within the support polygon of the robot’s foot ensures that the foot remains planted on the ground, assuming that friction is high enough to avoid slipping. The ZMP can be used to plan motion patterns that make the robot dynamically stable while walking. Controllers based on the ZMP criterion try to fol￾low a planned sequence of contact states and are often unable to change the landing positions in real time in response to lost contact. Current ZMP-based bipedal walking algorithms have difficulty handling unexpected perturbations, such as might be encountered with un￾even natural terrain (Fig. 56.9). Robots using ZMP differ from human locomotion in significant ways. For exam￾ple, unlike people, robots using ZMP typically do not exploit the natural dynamics of their legs, or control the impedance of their joints. Augmenting ZMP-based control is currently an ac￾tive area of research. As will be discussed in Sect. 56.5 on whole-body activities researchers are working to integrate manipulation and bipedal locomotion. For ex￾ample, when a robot walks while grasping a handrail, the contact could potentially increase the stability of the robot, but the ZMP criterion does not easily generalize Fig. 56.11 Example of controlled falling-down motion Fig. 56.12 The humanoid robot HRP-2P getting up from a lying￾down position to this task. So far, a generic and rigorous new criterion has not been established. As an example of an alternative mode of bipedal locomotion, some running robots have used controllers based on an inverted pendulum model to achieve sta￾ble gaits. These methods change the landing positions to keep the robot dynamically stable [56.37]. More re￾cently, researchers have begun to use the principles of bipedal passive-dynamic walkers to develop pow￾ered bipedal walkers that walk with high efficiency in a human-like way by exploiting natural dynam￾ics (Fig. 56.10 [56.36]). 56.3.2 Falling Down A human-scale robot should expect to fall from time to time in realistic conditions. A humanoid robot may fall down due to a large disturbance even if the motion is planned carefully and a sophisticated feedback con￾troller is applied to the robot. In this event, the robot could be damaged significantly during a fall, and could also damage the environment or injure people who are nearby. An important area of research is how to control the robot’s fall in order to gracefully recover or minimize damage. The Sony QRIO can control its falling motions in order to reduce the impact of touch down [56.38], al￾Part G 56.3

1314 Part G Human-Centered and Life-Like Robotics there is also the issue of getting back up again [56.40] Part G56. (Fig.56.12). 56.3.3 Sensing for Balance Rubber bushing w Bipedal walking needs to be robust to unexpected dis- turbances encountered during the execution of planned walking patterns.In these situations,walking can sometimes be stabilized with feedback control and 6-axis force sensor appropriate sensing.Many humanoid robots,such as Rubber sole Honda's Asimo,make use of accelerometers,gyro- Fig.56.13 Example of a humanoid foot structure for legged loco- scopes,and six-axis force/torque sensors to provide motion that uses compliance and force/torque sensing feedback to the robot during locomotion. Force/torque sensors have long been applied to ma- nipulators for the implementation of force control,but force/torque sensors with sufficient robustness to handle foot impact for a full-size humanoid robot are relatively new.When the foot of the robot touches down,the foot receives an impact which can disturb its walking.This impact can be rather large,especially when the robot is walking quickly.Some feet now incorporate a spring and damper mechanism as shown in Fig.56.13 in order to mitigate these problems.As with many other aspects of bipedal humanoid locomotion,foot design is currently an open problem. 56.3.4 Localization and Obstacle Detection In order for a humanoid robot to walk in unmod- Fig.56.14 Asimo and artificial landmarks on the floor eled environments.localization and obstacle detection are essential.Wheeled robots encounter similar is- though it is of a relatively small size (which simplifies sues while navigating,but full bipedal humanoids have the problem).Fujiwara et al.developed a falling mo- more-specialized requirements.For example,bipedal tion controller for a human-size humanoid robot that is humanoids have the ability to control contact with the falling backwards [56.39].Figure 56.11 shows an exam- world through their highly articulate legs. ple of a controlled falling motion.The general problem Artificial landmarks can simplify localization.As is still very much an active area of research.Similarly, shown in Fig.56.14,Honda's Asimo uses a camera mounted on its lower torso that looks down at the floor to find artificial markers for position correction [56.41]. Accurate positioning is important for long-distance nav- igation and stair climbing,since slippage usually occurs while walking and accumulated positional and direc- tional errors can lead to severe failures. Obstacle avoidance is also an important function for locomotion.Disparity images generated by stereo vision have been utilized for this purpose.For example,the plane segment finder [56.42]developed by Okada et al. helps detect traversable areas.Figure 56.15 shows the re- sult of detecting clear areas of the floor plane appropriate for gait generation. Humanoids require a great deal of computation Fig.56.15 Plane segment finder for detecting traversable floor area due to the need for sophisticated sensing and con-

1314 Part G Human-Centered and Life-Like Robotics 6-axis force sensor Rubber sole Rubber bushing Fig. 56.13 Example of a humanoid foot structure for legged loco￾motion that uses compliance and force/torque sensing Fig. 56.14 Asimo and artificial landmarks on the floor though it is of a relatively small size (which simplifies the problem). Fujiwara et al. developed a falling mo￾tion controller for a human-size humanoid robot that is falling backwards [56.39]. Figure 56.11 shows an exam￾ple of a controlled falling motion. The general problem is still very much an active area of research. Similarly, Fig. 56.15 Plane segment finder for detecting traversable floor area there is also the issue of getting back up again [56.40] (Fig. 56.12). 56.3.3 Sensing for Balance Bipedal walking needs to be robust to unexpected dis￾turbances encountered during the execution of planned walking patterns. In these situations, walking can sometimes be stabilized with feedback control and appropriate sensing. Many humanoid robots, such as Honda’s Asimo, make use of accelerometers, gyro￾scopes, and six-axis force/torque sensors to provide feedback to the robot during locomotion. Force/torque sensors have long been applied to ma￾nipulators for the implementation of force control, but force/torque sensors with sufficient robustness to handle foot impact for a full-size humanoid robot are relatively new. When the foot of the robot touches down, the foot receives an impact which can disturb its walking. This impact can be rather large, especially when the robot is walking quickly. Some feet now incorporate a spring and damper mechanism as shown in Fig. 56.13 in order to mitigate these problems. As with many other aspects of bipedal humanoid locomotion, foot design is currently an open problem. 56.3.4 Localization and Obstacle Detection In order for a humanoid robot to walk in unmod￾eled environments, localization and obstacle detection are essential. Wheeled robots encounter similar is￾sues while navigating, but full bipedal humanoids have more-specialized requirements. For example, bipedal humanoids have the ability to control contact with the world through their highly articulate legs. Artificial landmarks can simplify localization. As shown in Fig. 56.14, Honda’s Asimo uses a camera mounted on its lower torso that looks down at the floor to find artificial markers for position correction [56.41]. Accurate positioning is important for long-distance nav￾igation and stair climbing, since slippage usually occurs while walking and accumulated positional and direc￾tional errors can lead to severe failures. Obstacle avoidance is also an important function for locomotion. Disparity images generated by stereo vision have been utilized for this purpose. For example, the plane segment finder [56.42] developed by Okada et al. helps detect traversable areas. Figure 56.15 shows the re￾sult of detecting clear areas of the floor plane appropriate for gait generation. Humanoids require a great deal of computation due to the need for sophisticated sensing and con￾Part G 56.3

Humanoids 56.4 Manipulation 1315 trol.Customized computational hardware may help real time from the stereo cameras.This real-time mitigate this problem.For example,Sony's humanoid vision system has been used to detect floor ar- Part robot QRIO is equipped with a field-programmable eas,stair steps,and obstacles for navigation [56.43, 0 gate array (FPGA)to generate disparity maps in 44]. 3 56.4 Manipulation Hands and arms are the main interfaces with which one at the elbow,and three at the wrist.The use of seven humans act on the world around them.Manipulation DOFs results in a redundant degree of freedom with research within humanoid robotics typically focuses on respect to the six-DOF pose of the hand.To reduce me- the use of anthropomorphic arms,hands,and sensors chanical complexity,humanoid robot arms sometimes to perform tasks that are commonly performed by peo-have fewer than seven DOFs,for example,ARMAR- ple.Several chapters of the handbook relate to these III and Justin have seven-DOF arms,Cog and Domo goals,including Chap.24(Visual Servoing and Visual have six-DOF arms,and Asimo has five-DOF arms Tracking),Chap.26 (Motion for Manipulation Tasks),(Fig.56.16)[56.45,46]. and Chap.28(Grasping). Humanoid robot hands tend to vary more in their design (see Chap.15,Robot Hands).The human hand is 56.4.1 The Arm and Hand highly complex with over 20 DOFs(i.e.,approximately four DOFs per finger and a five-DOF thumb)in a very The kinematics of humanoid robot arms emulate the compact space with a compliant exterior,dense tactile human arm,which can be approximated by seven de-sensing,and muscular control.If a robot hand is to be grees of freedom(DOFs),with three at the shoulder, mounted on a robot arm,there are additional constraints in terms of the mass of the robot hand,since the hand sits at the end of the arm and must be efficiently moved in Fig.56.16 The humanoid robot Justin has two seven-DOF torque-controlled arms (DLR-Lightweight-Robot-III),and two 12-DOF hands (DLR-Hand-II).Justin's body is larger Fig.56.17 3-D object recognition by HRP-2 using versatile than a human's volumetric vision

Humanoids 56.4 Manipulation 1315 trol. Customized computational hardware may help mitigate this problem. For example, Sony’s humanoid robot QRIO is equipped with a field-programmable gate array (FPGA) to generate disparity maps in real time from the stereo cameras. This real-time vision system has been used to detect floor ar￾eas, stair steps, and obstacles for navigation [56.43, 44]. 56.4 Manipulation Hands and arms are the main interfaces with which humans act on the world around them. Manipulation research within humanoid robotics typically focuses on the use of anthropomorphic arms, hands, and sensors to perform tasks that are commonly performed by peo￾ple. Several chapters of the handbook relate to these goals, including Chap. 24 (Visual Servoing and Visual Tracking), Chap. 26 (Motion for Manipulation Tasks), and Chap. 28 (Grasping). 56.4.1 The Arm and Hand The kinematics of humanoid robot arms emulate the human arm, which can be approximated by seven de￾grees of freedom (DOFs), with three at the shoulder, Fig. 56.16 The humanoid robot Justin has two seven-DOF torque-controlled arms (DLR-Lightweight-Robot-III), and two 12-DOF hands (DLR-Hand-II). Justin’s body is larger than a human’s one at the elbow, and three at the wrist. The use of seven DOFs results in a redundant degree of freedom with respect to the six-DOF pose of the hand. To reduce me￾chanical complexity, humanoid robot arms sometimes have fewer than seven DOFs, for example, ARMAR￾III and Justin have seven-DOF arms, Cog and Domo have six-DOF arms, and Asimo has five-DOF arms (Fig. 56.16) [56.45, 46]. Humanoid robot hands tend to vary more in their design (see Chap. 15, Robot Hands). The human hand is highly complex with over 20 DOFs (i. e., approximately four DOFs per finger and a five-DOF thumb) in a very compact space with a compliant exterior, dense tactile sensing, and muscular control. If a robot hand is to be mounted on a robot arm, there are additional constraints in terms of the mass of the robot hand, since the hand sits at the end of the arm and must be efficiently moved in Fig. 56.17 3-D object recognition by HRP-2 using versatile volumetric vision Part G 56.4

1316 Part G Human-Centered and Life-Like Robotics nition and six-DOF pose estimation using models of the Part G56. objects to be manipulated (see Chap.23,3-D Vision and Recognition). An example of this approach is provided by the versatile volumetric vision (VVV)system,which is an edge-based 3-D vision system developed by Tomita et al.[56.54].The VVV system has been utilized to find Hot-glue gun Screwdriver Bottle and grasp objects during everyday manipulation tasks. Fig.56.18 Using visual motion,Domo detects (white Figure 56.17 shows a demonstration of the HRP-2 recog- cross)the tips of tool-like objects it is rigidly grasping nizing a beverage can on a table so that it can pick up the with error (white circle)comparable to the error achieved can and throw it into a trash can.This integrated system with hand labels(black cross) was demonstrated at Aichi EXPO 2005,where it en- abled a human operator to control the HRP-2 humanoid space.Researchers have approximated the human hand robot in a semi-autonomous fashion with reduced ef- with varying levels of accuracy,including the ACT hand, fort [56.55].The VVV system has also helped HRP-2 the 20-DOF Shadow Hand.the 12-DOF DLR-Hand-II. carry a table in cooperation with a human [56.561. the 11-DOF Robonaut hand,the two-DOF Cog hand, Other robots have used similar approaches.For ex- and the one-DOF Asimo hand [56.47-501. ample,Robonaut from the National Aeronautics and The ACT hand is an excellent representative of the Space Administration (NASA)Johnson Space Center high-fidelity end of the spectrum,since it emulates the (JSC)has used model-based vision to perform every- bone structure,inertial properties,and actuation of the day manipulation tasks,such as tightening lug nuts on human hand in addition to the kinematics.Humanoid a wheel [56.57]. robot hands often include passive degrees of freedom. For example,one DOF of the two-DOF Cog hand Feature-Based Vision controlled a multijointed power grasp,while the other Another approach to visual perception for humanoid controlled a multijointed two-fingered precision grasp. robots is feature-based vision.Encoding tasks in terms Studies indicate that many human grasps can be ap- of task-relevant features,such as the tip of a tool or proximated with two degrees of freedom [56.53],so the contact surface of a hand.offers an alternative to simplified hands may be sufficient to emulate a variety approaches that use detailed 3-D models of objects. of human manipulation activities. In order to generalize a task across different objects, only the task-relevant features need to be detected and 56.4.2 Sensing for Manipulation mapped. An example of this approach is provided by the hu- Model-Based Vision manoid robot Domo,which uses task-relevant features A common approach to visual perception for humanoid to perform everyday tasks such as pouring,stirring,and robots is real-time three-dimensional (3-D)object recog- brushing.Domo detects features such as the opening of a container or the tip of a tool (Fig.56.18)and then visu- ally servoes these features with respect to one another in order to perform a task.Once the objects are in contact with one another,Domo uses force sensing and compli- ance to simplify tasks such as regrasping,inserting,and placing objects [56.52,58]. Active Perception Through action,robots can simplify perception.Hu- manoid robots have used this approach in ways that Fig.56.19a,b The humanoid robots Obrero (a)and Domo(b)use are reminiscent of human behavior.For example,a hu- passive compliance and force control to safely reach out into the manoid robot can reach out into the world to physically world.Obrero haptically grasps an object [56.51].Domo physi- sense its surroundings(Fig.56.19),or induce visual mo- cally finds the shelf,and uses force control to let objects settle into tion through physical contact so as to better estimate the place [56.52] extent of manipulable objects [56.59].Similarly,a hu-

1316 Part G Human-Centered and Life-Like Robotics Hot-glue gun Screwdriver Bottle Fig. 56.18 Using visual motion, Domo detects (white cross) the tips of tool-like objects it is rigidly grasping with error (white circle) comparable to the error achieved with hand labels (black cross) space. Researchers have approximated the human hand with varying levels of accuracy, including the ACT hand, the 20-DOF Shadow Hand, the 12-DOF DLR-Hand-II, the 11-DOF Robonaut hand, the two-DOF Cog hand, and the one-DOF Asimo hand [56.47–50]. The ACT hand is an excellent representative of the high-fidelity end of the spectrum, since it emulates the bone structure, inertial properties, and actuation of the human hand in addition to the kinematics. Humanoid robot hands often include passive degrees of freedom. For example, one DOF of the two-DOF Cog hand controlled a multijointed power grasp, while the other controlled a multijointed two-fingered precision grasp. Studies indicate that many human grasps can be ap￾proximated with two degrees of freedom [56.53], so simplified hands may be sufficient to emulate a variety of human manipulation activities. 56.4.2 Sensing for Manipulation Model-Based Vision A common approach to visual perception for humanoid robots is real-time three-dimensional (3-D) object recog￾a) b) Fig. 56.19a,b The humanoid robots Obrero (a) and Domo (b) use passive compliance and force control to safely reach out into the world. Obrero haptically grasps an object [56.51]. Domo physi￾cally finds the shelf, and uses force control to let objects settle into place [56.52] nition and six-DOF pose estimation using models of the objects to be manipulated (see Chap. 23, 3-D Vision and Recognition). An example of this approach is provided by the versatile volumetric vision (VVV) system, which is an edge-based 3-D vision system developed by Tomita et al. [56.54]. The VVV system has been utilized to find and grasp objects during everyday manipulation tasks. Figure 56.17 shows a demonstration of the HRP-2 recog￾nizing a beverage can on a table so that it can pick up the can and throw it into a trash can. This integrated system was demonstrated at Aichi EXPO 2005, where it en￾abled a human operator to control the HRP-2 humanoid robot in a semi-autonomous fashion with reduced ef￾fort [56.55]. The VVV system has also helped HRP-2 carry a table in cooperation with a human [56.56]. Other robots have used similar approaches. For ex￾ample, Robonaut from the National Aeronautics and Space Administration (NASA) Johnson Space Center (JSC) has used model-based vision to perform every￾day manipulation tasks, such as tightening lug nuts on a wheel [56.57]. Feature-Based Vision Another approach to visual perception for humanoid robots is feature-based vision. Encoding tasks in terms of task-relevant features, such as the tip of a tool or the contact surface of a hand, offers an alternative to approaches that use detailed 3-D models of objects. In order to generalize a task across different objects, only the task-relevant features need to be detected and mapped. An example of this approach is provided by the hu￾manoid robot Domo, which uses task-relevant features to perform everyday tasks such as pouring, stirring, and brushing. Domo detects features such as the opening of a container or the tip of a tool (Fig. 56.18) and then visu￾ally servoes these features with respect to one another in order to perform a task. Once the objects are in contact with one another, Domo uses force sensing and compli￾ance to simplify tasks such as regrasping, inserting, and placing objects [56.52, 58]. Active Perception Through action, robots can simplify perception. Hu￾manoid robots have used this approach in ways that are reminiscent of human behavior. For example, a hu￾manoid robot can reach out into the world to physically sense its surroundings (Fig. 56.19), or induce visual mo￾tion through physical contact so as to better estimate the extent of manipulable objects [56.59]. Similarly, a hu￾Part G 56.4