Cybernetic Human HRP-4C: a Humanoid Robot with Human-Like Proportions

Cybernetic Human HRP-4C: A humanoid robot with human-like proportions

Shuuji KAJITA, Kenji KANEKO, Fumio KANEIRO, Kensuke HARADA, Mitsuharu MORISAWA, Shin’ichiro NAKAOKA, Kanako MIURA, Kiyoshi FUJIWARA, Ee Sian NEO, Isao HARA, Kazuhito YOKOI, Hirohisa HIRUKAWA

Abstract Cybernetic human HRP-4C is a humanoid robot whose body dimensions were designed to match the average Japanese young female. In this paper, we ex- plain the aim of the development, realization of human-like shape and dimensions, research to realize human-like motion and interactions using speech recognition.

1 Introduction

Cybernetics studies the dynamics of information as a common principle of com- plex systems which have goals or purposes. The systems can be machines, animals or a social systems, therefore, cybernetics is multidiciplinary from its nature. Since Norbert Wiener advocated the concept in his book in 1948[1], the term has widely spreaded into academic and pop culture. At present, cybernetics has diverged into robotics, control theory, artificial intelligence and many other research fields, how- ever, the original unified concept has not yet lost its glory. Robotics is one of the biggest streams that branched out from cybernetics, and its goal is to create a useful system by combining mechanical devices with information technology. From a practical point of view, a robot does not have to be humanoid; nevertheless we believe the concept of cybernetics can justify the research of humanoid robots for it can be an effective hub of multidiciplinary research. WABOT-1, the world first humanoid robot developed by Kato and his colleagues in 1973[2], was built as an integrated system having two arms, two legs, a voice

Shuuji Kajita, Kenji Kaneko, Fumio Kanehiro, Kensuke Harada, Mitsuharu Morisawa, Shin’ichiro Nakaoka, Kanako Miura, Kiyoshi Fujiwara, Ee Sian Neo, Isao Hara, Kazuhito Yokoi, Hirohisa Hirukawa AIST, 1-1-1 Umezono, Tukuba, Ibaraki, Japan e-mail: {s.kajita,k.kaneko,f- kanehiro,kensuke.harada,m.morisawa,s.nakaoka,kanako.miura,k-fujiwara,rio.neo,isao- hara,Kazuhito.Yokoi,hiro.hirukawa}@aist.go.jp

1 2 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al. recognition system etc. Later, their group developed a piano playing humanoid, WABOT-2 in 1985[3] In the early 1990s, Brooks and his colleague started to build humanoid robots as physical embodiment of artiﬁcial intelligence[4]. Research activities in ATR[5] and the RobotCub project[6] can be considered to have the same intention, namely humanoid robotics for cognitive science. Since the breakthrough on biped walking done by Hirai et al.[7], many research projects on biped humanoid robots were conducted. Research was carried on generation of walking motion[8], jogging and running [9, 10], efﬁcient walking[11], human-like walking[12], dynamic whole body balance control[13] and so forth. There also exists research working on facial expression of humanoid robots [14, 15, 16, 17]. These works target social communication and interaction, which is another aspect of cybernetics. In AIST, we have been developing a series of humanoid robots[19, 20]. Cyber- netic human HRP-4C is our latest development, which is a humanoid robot designed to have body dimensions close to average Japanese young female(Fig.1, Table 1). This paper explains the goal of our project and introduce the developed robot system.

Fig. 1 Cybernetic human HRP-4C Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 3

Table 1 Principal speciﬁcations of HRP-4C Height 1,580 [mm] Weight(with batteries) 43 [kg] Total DOF 42 DOF Face 8 DOF Neck 3 DOF Arm 6 DOF × 2 Hand 2 DOF × 2 Waist 3 DOF Leg 6 DOF × 2 CPUs Motion controller Intel Pentium M 1.6GHz Speech recognition VIA C7 1.0GHz Sensors Body Posture sensor Sole 6-axies force sensor × 2 Batteries NiMH

2 Aim of the development

2.1 User Centered Robot Open Architecture

Fig. 2 User Centered Robot Open Architecture (UCROA)

HRP-4C was developed in the User Centered Robot Open Architecture (UCROA) Project which is one of the projects under the AIST Industrial Transformation Re- search Initiative, a 3-year industry-academia joint project implemented by AIST from the ﬁscal year of 2006. The goal of the UCROA is to show society that it is pos- 4 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al.

sible to develop robot products that meet the speciﬁcations required by users combining reusable core technologies and developing prototype next-generation robots which can be used in practice by 2010 and for which the market is expected to be large. Three prototype robots were developed in the project (Fig.2). 1. Logistics support robot 2. Personal service robot 3. Cybernetic human For the logistics support robot, we developed a robotic system for warehouses. For the personal service robot, a small manipulator system to assist daily life of hand- icapped people was developed. The third subject of the UCROA is the cybernetic human, whose concept is explained in the next subsection.

2.2 Concept of Cybernetic human

A humanoid robots attract many people, because of its human-like appearance and behavior. Although a robot with very human-like appearance is called an “android” in general, we coined a new term Cybernetic human to deﬁne a humanoid robot with the following features. 1. Have the appearance and shape of a human being 2. Can walk and move like a human being 3. Can interact with humans using speech recognition and so forth Such robots can be used in the entertainment industry, for example, exhibitions and fashion shows. It can be also used as a human simulator to evaluate devices for humans. As the successor of our previous humanoid robots HRP-2 and HRP-3[19, 20], we call our new humanoid robot HRP-4C, “C” stands for cybernetic human.

3 Realization of human-like shape and dimensions

3.1 Target speciﬁcations

To determine the target shape and dimensions of HRP-4C, we used the anthropometric database for Japanese population, which was measured and compiled by Kouchi et al.[18]. The database provides the dimensions of the following four different Japanese groups. 1. Young male: aged 19-27, average 20.5 years old, 110 samples 2. Young female: aged 19-27, average 20.2 years old, 107 samples 3. Aged male: aged 60-82, average 68.6 years old, 51 samples Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 5

4. Aged female: aged 60-80, average 66.9 years old, 50 samples Considering the entertainment applications like fashion show, we picked the average young female data. Figure 3 shows part of the dimensions provided by the database.

Upper arm circumference: 253 Forearm circumference: 228 Head breadth: 156

Chest circumference: 830

Waist circumference: 659

Head length: 181 Stature: 1586 Total head height: 231 Total Suprasternal height: 1281 Waist height: 993 Waist Cristal height: 944 Iliac spine height: 850 Trochanterion height: 801 Trochanterion Symphyseal height: 789 Lateral epicondyle height: 429

Thigh circumference: 543 [mm] Calf circumference: 351

Fig. 3 Anthropometric data of average young Japanese female [18]

In the early stage of design, we explored the possible choice and arrangement of mechanical and electric devices by scaling up and down Fig.3 to have the stature between 145cm and 160cm. In our ﬁnal design, HRP-4C is 158cm tall which is very close to the average of the young female, 158.6cm.

3.2 Joint conﬁguration

To obtain graceful motion of females, we asked a professional walking model to perform walking, turning, sitting on chair, and other motions. The positions of 86 markers attached to her body were captured by Vicon Motion Systems, a 3D optical motion capture device (Fig.4). This data was used to evaluate the different joint conﬁgurations proposed for HRP-4C structure. By calculating joint angles to realize the captured motion, the necessary movable range was estimated. In addition, we 6 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al.

Fig. 4 Motion capturing of professional walking model (Walking Studio Rei)

estimated the motor power during biped walking to determine the appropriate leg joint conﬁguration.

(a)

(b)

Fig. 5 Joint conﬁguration of HRP-4C body (The head and hands are omitted)

Figure 5 shows the joint conﬁguration we ﬁnally decided on for HRP-4C. In this drawing, the joints for head and hands are omitted. It has the following characteristic compared with our former humanoid robots, HRP-2 and HRP-3. Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 7

• Roll axes of waist and neck were added to realize human-like behavior (Fig.5 arrow (a)) • Slanted links of forearm and thigh to mimic human (Fig.5 arrow (b)) • Use of standard hip joint structure to realize natural waist line (For HRP-2 and HRP-3, we used cantilever type hip joint[19, 20])

3.3 Design of the body mechanism

Figure 6 shows the designed body mechanism of HRP-4C (right) and HRP-2 (left). By comparison, we see that HRP-4C realized much smaller chest and hip as well as slender extremities. To realize this body mechanism, we adopted the following technologies. • PCI-104 single board computer and peripheral boards for the whole body motion control • Distributed network motor drivers • Development of slim ankle mechanism For further details of the body mechanism and electronics, see our next report[21]. 1580

Design: Kawada Industries. Inc.

Fig. 6 HRP-2 and mechanism of HRP-4C 8 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al. 3.4 Design of appearance

After the mechanical design of the leg part of HRP-4C was completed, we started to consider the design of appearance. Figure 7 shows the proposed appearances of HRP-4C. Fig.7(a) is a metallic robot with womanly form, (b) is a human-like design with artiﬁcial skin and a dress to cover the mechanical parts, (c) is a conventional humanoid robot design and (d) is a design as a mannequin.

(a) (b) (c) (d)

Fig. 7 Proposed designs for HRP-4C

One of the problems of the design like Fig.7(b) is when the degree of similarity passes a certain level, it can make people feel strange or even fearful. This effect was named the uncanny valley by Mori[22]. He also pointed out that it can be ampliﬁed by the difference of motion pattern between the robot and human. This is serious for an entertainment robot. Considering the uncanny valley and the impact as the entertainment robot, the ﬁnal design was decided to be between (a) and (b), that is metallic robot body with a human-like face (Fig.8). We also decided to make the robot face not so real. Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 9

Fig. 8 Final design for HRP-4C

3.5 Head and hands

To realize a head size close to the average young female (Fig.3) and achieve weight reduction, we limited the numbers of degrees of freedom (DOF) of the head. As the minimum DOF to create facial expression we chose the movements shown in Table 2. Since each pair of the eyebrows, eyelids, eyeballs pan and eyeballs tilt are actuated by one servomotor, HRP-4C cannot perform some facial expressions like winking. Figure 9 shows the face of HRP-4C. Despite the above noted limitations, HRP-4C can perform effective facial expressions like smile, surprise, anger etc.

Table 2 Head joints of HRP-4C Joint name DOF Joint name DOF Eyebrows 1 Mouth 1 Eyelids 1 Upper lip 1 Eyeballs pan 1 Lower lip 1 Eyeballs tilt 1 Cheek 1 Total 8 10 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al.

Fig. 9 Face of HRP-4C

We designed the hand of HRP-4C to create motion which is used in dance performance. Its DOF was again limited to the minimum (Table 3). The four ﬁngers from the index to the little ﬁngers are driven by one servomotor and the thumb is driven by another. The size of the hand became bigger than the average young female, due to the selected servomotors.

Table 3 Hand joints of HRP-4C Joint name DOF Index, middle, third and little ﬁngers 1 Thumb 1 Total 2

4 Towards human-like motion and walking

We developed a couple of algorithms to generate human-like motion and walking from the mocap data obtained in 3.2[23, 24]. Figure 10 shows HRP-4C performing a 90 degree turn which was created from the captured human motion. We developed a new control software to stabilize a humanoid robot motion with stretched knee. Currently, the reliability of the controller is not enough, and we are improving it. We also have developed a software tool to manually program HRP-4C. It can be used to quickly create the motion of HRP-4C through an interactive user interface [25]. Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 11

t=3.1[s] t=4.1[s]

t=5.05[s] t=6.16[s]

Fig. 10 90 degree turn based on captured motion

5 Interaction using speech recognition

We used the open source speech recognition engine Julian[26]. It runs on the CPU embedded in the head of HRP-4C. To realize robust recognition against ambient noise, an operator uses a wireless Bluetooth microphone to send voice commands. The recognition result is transmitted to the motion control software running on the other CPU in the body of HRP-4C via RT middleware[27]. Figure 11 shows an operator speaking the voice command “Look surprised!” and HRP-4C demonstrat- ing the surprise motion. 12 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al.

Fig. 11 HRP-4C responding to the voice command ”Look surprised!”

6 Conclusions and future work

In this paper, we gave an overview of the development of our new humanoid robot, cybernetic human HRP-4C. The robot has the appearance and shape of a human being, speciﬁcally, an average young Japanese female. It can perform biped walk using its own battery and it can interact with humans using speech recognition. The software of HRP-4C is still under development for the project was carried out with a tight schedule. One of the urgent goal is to realize a reliable human-like biped walking with stretched knees.

Acknowledgments

We would like to thank Dr. Junji Ito, director of AIST, for his generous help ob- taining the ﬁnancial support to the project. Moreover, we also thank Dr. Masaahi Mochimaru, Dr. Kei Aoki and other members of the Digital Human Research Cen- ter, who kindly helped our motion capturing experiment. Without the help of above mentioned people, the development of HRP-4C was not possible. Cybernetic Human HRP-4C: A humanoid robot with human-like proportions 13 References

1. Norbert Wiener, Cybernetics: or Control and Communication in the Animal and the Machine, MIT Press, 1948. 2. Ichiro Kato, “Development of biped walking robot WABOT-1,” Biomechanism 2, pp.173-214, 1973 (in Japanese). 3. S.Sugano, I.Kato, “WABOT-2: Autonomous robot with Dexterous Finger-Arm — Finger- Arm Coordination Control in Keyboard Performance —,” Proc. of IEEE Int. Conference on Robotics and Automation, vol.4, pp.90-97, 1987. 4. R.A.Brooks, C.Breazeal, M.Marjanovic, B.Scassellati and M.Williamson, “The Cog Project: Building a Humanoid Robot,” in C. Nehaniv, ed., Computation for Metaphors, Analogy and Agents, Vol. 1562 of Springer Lecture Notes in Artificial Intelligence, Springer-Verlag, 1998. 5. J.Peters, S.Vijayakumar, S.Schaal, “Reinforcement Learning for Humanoid Robotics,” Proc. on IEEE-RAS Int. Conference on Humanoid Robots, 2003. 6. Metta, G., Sandini, G., Vernon, D., Natale, L., Nori, F. , “The iCub humanoid robot: an open platform for research in embodied cognition,” In PerMIS: Performance Metrics for Intelligent Systems Workshop, 2008. 7. K.Hirai, M.Hirose, Y.Haikawa and T.Takenaka, “The Development of Honda Humanoid Robot,” Proc. IEEE Int. Conference on Robotics and Automation, pp.1321-1326, 1998. 8. S.Kagami, K.Nishiwaki, T.Kitagawa, T.Sugihara, M.Inaba and H.Inoue “A Fast Generation Method of a Dynamically Stable Humanoid Robot Trajectory with Enhanced ZMP Con- straint,” in Proc. of IEEE Int. Conf. on Humanoid Robotics, 2000. 9. S.Lohmeier, T.Bushmann and H.Ulbrich, “Humanoid Robot LOLA,” Proc. IEEE Int. Confer- ence on Robotics and Automation, pp.775-780, 2009. 10. R.Tajima, D.Honda, and K.Suga, “Fast Running Experiments Involving a Humanoid Robot,” Proc. of IEEE Int. Conference on Robotics and Automation, pp.1571-1576, 2009. 11. S.Collins, A.Ruina, R.Tedrake, M.Wisse, “Efficient Bipedal Robots Based on Passive-Dynaic Walkers,” Science, vol.307, pp.1082-1085, 2005. 12. Y.Ogura, H.Aikawa, K.Shimomura, H.Kondo, A.Morishima, H.Lim and A.Takanishi, “Devel- opment of A Humanoid Robot WABIAN-2,” Proc. of IEEE Int. Conference on Robotics and Automation, pp.76-81, 2006. 13. S.H.Hyon, R.Osu and Y.Otaka, “Integration of multi-level postural balancing on humanoid robots,” Proc. IEEE Int. Conference on Robotics and Automation, pp.1549-1556, 2009. 14. C.Breazeal and B.Scasselati, “How to build robots that make friends and influence people,” in Proc. of IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp.858-863, 1999. 15. H. Ishiguro, “Android science: conscious and subconscious recognition,” Connection Science, vol.18, no.4, pp.319-332, 2006. 16. J.H.Oh, D.Hanson, W.S.Kim, I.Y.Han, J.Y.Kima and I.W.Park, “Design of android type humanoid robot albert HUBO,” in Proc. of the IEEE/RSJ Inter. Conf. on Intelligent Robots and systems, pp.1428-1433, 2006. 17. T.Hashimoto, S.Hiramatsu and H.Kobayashi, “Dynamic display of facial expressions on the face robot made by using a life mask,” in Proc. IEEE-RAS Inter . Conf. on Humanoid Robots, pp.521-526, 2008. 18. M.Kouchi, M.Mochimaru, H.Iwasawa and S.Mitani, “Anthropometric database for Japanese Population 1997-98,” Japanese Industrial Standards Center (AIST, MITI), http://riodb.ibase.aist.go.jp/dhbodydb/ , 2000. 19. K.Kaneko, F.Kanehiro, S.Kajita, H.Hirukawa, T.Kawasaki, M.Hirata, K.Akachi and T.Isozumi ,“Humanoid Robot HRP-2,” Proc. IEEE Int. Conference on Robotics and Automa- tion, pp.1083-1090, 2004. 20. K.Kaneko, K.Harada, F.Kanehiro, G.Miyamori and K.Akachi, “Humanoid Robot HRP-3” Proc. IEEE/RSJ Int. Conference on Intelligent Robots and Systems, pp.2471-2478, 2008. 21. K.Kaneko, F.Kanehiro, M.Morisawa, K.Miura, S.Nakaoka and S.Kajita, “Cybernetic Human HRP-4C,” 9th IEEE-RAS Int. Conference on Humanoid Robots, 2009 (submitted paper). 14 Kajita, Kaneko, Kanehiro, Harada, Morisawa, Nakaoka, Miura, Fujiwara et al.

22. M.Mori: “The uncanny valley” Energy, vol.7, no.4, pp.33-35, 1970. [On- line] http://www.androidscience.com/theuncannyvalley/ proceedings2005/uncannyvalley.html 23. K.Harada, K.Miura, M.Morisawa, K.Kaneko, S.Nakaoka, F.Kanehiro, T.Tsuji, S.Kajita: “To- ward Human-Like Walking Pattern Generator,” Proc. IEEE/RSJ Int. Conference on Intelligent Robots and Systems, 2009 (to appear). 24. K.Miura, M.Moisawa, S.Nakaoka, F.Kanehiro, K.Harada, K.Kaneko, S.Kajita: “Towards Human-like Locomotion of Humanoid Robots — Walking and Turning based on Motion Cap- ture Data —,” 9th IEEE-RAS Int. Conference on Humanoid Robots, 2009 (submitted paper). 25. Nakaoka, Kanehiro, Miura, Morisawa et al.“Creating Facial Motions of Cybernetic Human HRP-4C,” 9th IEEE-RAS Inter. Conf. on Humanoid Robots, 2009 (submitted paper). 26. A. Lee, T. Kawahara and K. Shikano. ”Julius — an open source real-time large vocabulary recognition engine.” In Proc. European Conference on Speech Communication and Technol- ogy (EUROSPEECH), pp. 1691–1694, 2001. 27. OpenRTM-aist,http://www.is.aist.go.jp/rt/OpenRTM-aist, 2007.