Experimental Study of Humanoid Robot HRP-1S

The International Journal of Robotics Research http://ijr.sagepub.com

Experimental Study of Humanoid Robot HRP-1S Kazuhito Yokoi, Fumio Kanehiro, Kenji Kaneko, Shuuji Kajita, Kiyoshi Fujiwara and Hirohisa Hirukawa The International Journal of Robotics Research 2004; 23; 351 DOI: 10.1177/0278364904042194 The online version of this article can be found at: http://ijr.sagepub.com/cgi/content/abstract/23/4-5/351

Published by: http://www.sagepublications.com

On behalf of:

Multimedia Archives

Additional services and information for The International Journal of Robotics Research can be found at: Email Alerts: http://ijr.sagepub.com/cgi/alerts Subscriptions: http://ijr.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations (this article cites 2 articles hosted on the SAGE Journals Online and HighWire Press platforms): http://ijr.sagepub.com/cgi/content/refs/23/4-5/351

Downloaded from http://ijr.sagepub.com at PENNSYLVANIA STATE UNIV on February 12, 2008 © 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

Experimental Study of Humanoid Robot HRP-1S

Kazuhito Yokoi Fumio Kanehiro Kenji Kaneko Shuuji Kajita Kiyoshi Fujiwara Hirohisa Hirukawa Intelligent Systems Institute National Institute of Advanced Industrial Science and Technology (AIST) 1-1-1 Umezono, Tsukuba, Ibaraki, 305-8568, Japan [email protected]

Abstract We have developed a humanoid robot HRP-1S that is capable of simultaneous whole body motion control. In phase one of the Humanoid Robotics Project (HRP) of the Ministry of Economy, Trade and Industry of Japan, Honda R&D Co. Ltd has produced humanoid robot HRP-1 as a humanoid research platform to ease the development of good applications for humanoid robots in phase two of HRP. However, HRP-1 controls its legs and arms separately, and is not suitable for some applications. We modified the control hardware of HRP-1 and implemented our own control software. The modified robot was designated HRP-1S. The motion controller of HRP-1S is built in accordance with the plug-in architecture. We have developed some plug-ins including a real-time walking pattern generator (KWALK plug-in) and a reflex controller (STABILIZER plug-in) to maintain the dynamic balance of HRP-1S. In this paper, we present the system architecture of the control system of HRP-1S and several experimental results.

KEY WORDS—humanoid robot, Humanoid Robotics Project (HRP), plug-in architecture, control system

1. Introduction The humanoid robot has always been our dream. The first full-size humanoid robot “WABOT-I” was created in 1972 (Kato et al. 1972). In the following 20 years, the subject of many researchers was just biped walking (Takanishi et al. 1985; McGeer 1990; Zheng and Shen 1990; Kajita and Tani 1991). Inaba (1993) hammered out “Remote-Brain Robotics” and started to build small-size humanoid robots. The robots The International Journal of Robotics Research Vol. 23, No. 4–5, April–May 2004, pp. 351-362, DOI: 10.1177/0278364904042194 ©2004 Sage Publications

could not only perform biped walking but also many types of whole-body behavior such as swing, carry a object, roll over, use a mobile base and so on (Kanehiro et al. 2000). However, most people were still skeptical about the realization of a full-sized humanoid robot. The most impressive humanoid robot, Honda P2, came about in 1996 (Hirai 1997). This was a turning point, and research on humanoid robots has become very active. Not only academia (Yamaguchi and Soga 1999; Espiau and Sardain 2000; Konno et al. 2000; Yamasaki et al. 2000; Furuta et al. 2001; Gienger, Löffler, and Pfeiffer 2001; Kagami et al. 2001; Nagakubo, Kuniyoshi, and Cheng 2001; Wolff and Nordin 2001; Lorch et al. 2002), but also several industries produce humanoid robots (Hirose et al. 2001; Ishida and Kuroki 2001): FREEDOM, BestTechnology Co., Ltd (http://www.besttechnology.co.jp/), Miniature Humanoid Robot “HOAP1”, Fujitsu Automation Ltd (http://www.automation.fujitsu.com/en/products/products07. html), Humanoid Robot “isamu”, Kawada Industries, Inc. (http://www.kawada.co.jp/ams/isamu/index_e.html) and Robos Humanoid Robot “KOZOH”, robos (http://www.robos.co. jp/index_pro.html). The application area of humanoid robots has, however, still been limited to amusement, entertainment, as a research tool, or a companion. The Ministry of Economy, Trade and Industry (METI) of Japan has run R&D projects on humanoid robotics since FY 1998 (Inoue et al. 2001). The project is run on a new scheme, called a platform-based approach, in which a platform is developed at phase one and it is utilized by contributors of the project as an infrastructure for R&D in phase two. The main mission of phase two is finding a promising application of humanoid robots that can initialize a mass production spiral for the humanoid robot industry. Honda R&D Co. Ltd has produced the humanoid robot platform HRP-1 in phase one of the Humanoid Robotics

351 Downloaded from http://ijr.sagepub.com at PENNSYLVANIA STATE UNIV on February 12, 2008 © 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

352

THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / April–May 2004

Wireless LAN 2Mbps

Robot Platform HRP-1S CPU Board(Body Control) Intel PentiumIII 800MHz 512MB Memory 128MB Compact Flash

Honda I/O Board Wired LAN 100Mbps Internet

Fig. 1. Minimum set of control hardware system of HRP-1S.

Project (HRP; Inoue et al. 2000a). The legs and arms of HRP1 must be controlled separately. So, HRP-1 must stop when it wants to use its arms. Additionally, HRP-1 cannot change its body posture at the user level. Only two applications, “maintenance tasks of industrial plants” and “security services of home and office”, of HRP can be developed on HRP-1. The other two application areas, “human care” and “teleoperation of construction machines”, require simultaneous operation of its arms and legs. In order to satisfy this demand, we have developed a humanoid robot HRP-1S capable of whole-body motion control. HRP-1S is made by replacing the main control CPU and the software of HRP-1. The rest of the paper is organized as follows. In Section 2, we present the hardware specification of HRP-1S. The control software of HRP-1S is described in Section 3. Section 4 shows experimental results. We conclude the paper in Section 5.

Table 1. Specification of the Humanoid Robot Dimensions H 1600 × W 600 × D 595 mm Weight 99 + 17 kg (battery) Power supply Line or battery, 136 V, 50 A max, 1 kW when walking Weight 99 + 17 kg (battery) Working Office/laboratory, 10–35◦ C environment Head Neck: 1 DoF, Pan +/ − 60◦ Two color CCD camera: stereo, 1 DoF Tilt +85◦ −15◦ Arms 7 DoF/arm (shoulder 3, elbow 1, wrist 3) Upper arm length 280 mm, lower arm length 280 mm Wrist force/torque sensor Hands Load 2 kgf max Legs 6 DoF/leg (hip 3, knee 1, ankle 2) Ankle force/torque sensor Sensors Gyroscope, G-force sensor Actuators Brushless DC servo motor, harmonic drive gear

sensor measurements are sent to the CPU through the Honda I/O board. Each joint servo control is performed by the distributed motor driver. It is possible to mount extra I/O boards if required.

2. Humanoid Robot HRP-1S

3. Motion Control Systems of HRP-1S

The dimensions of HRP-1S are height 1600 mm, width 600 mm, and weight 99 kg excluding batteries. The mechanical hardware is identical to HRP-1 (Inoue et al. 2000a). It has six degrees of freedom (DoF) in each leg and eight DoF in each arm including hands with 1-DoF grippers. Each joint is actuated by a brushless DC servo motor with a harmonic-drive reduction gear. Brushless DC servo amplifiers, wireless Ethernet are embedded in the body. The body is equipped with an inclination sensor that consists of gyroscopes and G-force sensors. Each foot and wrist is equipped with a force/torque sensor. Two video cameras are mounted in the head. The robot can use a Ni–Zn battery and operate for more than 30 min with the battery. Figure 1 shows the minimum set of the control hardware system of HRP-1S. It consists of a CPU board and Honda I/O board. The CPU board is VMIVME-7740-877 from VMIC, Single Board Computers (http://www.vmic.com/products/ embeddedpc/products/). The Honda I/O board handles input from all types of sensors used in the robot and sends output to motor drivers. It just relays the data and has no control software for the robot. All joint motion commands and all

Figure 2 shows the architecture of the motion control system of HRP-1S. The real-time controller for whole-body movements runs on the CPU board of HRP-1S. The operating system used is ART-Linux (Ishiwata and Matsui 1998). ARTLinux enables the execution of real-time processes at the user level so that users can implement real-time applications in the same way as non-real-time ones. This feature of ART-Linux is very helpful because it is possible to use the identical controller for the simulator as well as the real robot (Kanehiro et al. 2001a). In recent years, we have seen commercial software providing function expansion by using a plug-in architecture. Using this architecture, users and third parties are able to develop their own expansion modules. Additionally, the software can have a simple, compact, and highly expandable structure. This software structure is very effective as the software system of a humanoid robot. The humanoid robot can be versatile but it is extremely difficult to prepare all the functions from the beginning. A software architecture that can supply incremental expansion of functions must be introduced. The motion controller is, therefore, built as a plug-in architecture (Kanehiro et al. 2001b), where application software is


Yokoi et al. / Humanoid Robot HRP-1S

Whole Body Controller

Realtime Thread

Non-realtime Thread

Inside HRP-1S

353

Plug-in Manager Status Monitor I/O Library

Shared Memory Clone

Adaptor

Command

I/O Library

Command Interpreter

Plugin Plugin Plugin Plugin Plugin

Output Motor Command, Input Sensory Information

Adaptor

Shared Memory On Demand

Honda I/O Board

Joint Servo Driver

Cyclic

Fig. 3. Plug-in manager.

Fig. 2. Architecture of motion control system of HRP-1S.

separated into two layers at the adapter level. We call these the I/O library and plug-in manager, as shown in Figure 2. 3.1. I/O Library The I/O Library of HRP-1S handles input and output to shared memory that is used for achieving real-time communication between the Honda I/O board and the control CPU board. All motion commands and measurements of the robot are written to the shared memory every 5 ms, therefore we can monitor the internal state of the robot by cloning the image of the shared memory at fixed time intervals. The control CPU gets sensor data from the Honda I/O board and sends motion commands to the Honda I/O board via the shared memory every 5 ms. The Honda I/O board just distributes the motion commands to the motor drivers and the motor driver then controls each joint angle. 3.2. Adapter The controller API of OpenHRP (Hirukawa et al. 2001; Kanehiro et al. 2002) is not identical to that of the Honda I/O board. We introduced a software Adapter whose API is the abstracted controller API of the Honda I/O board in order to develop an identical controller for HRP-1S and its virtual counterpart. Thanks to this unification, the controller can share software with the dynamics simulation of OpenHRP, including the parameter parser, kinematics and dynamics computations and the collision detector. At the same time, we can run a simulation using the exact same code as is used on the real robot. This helps us to cut out the non-trivial porting that may induce trouble when the code run in the simulation is adapted to the real robot. 3.3. Plug-in Manager The plug-in manager serves as the backbone for the control system of HRP-1S. Each element of the control system is

treated as a plug-in. The plug-in manager loads a suitable plugin definition file for a task for HRP-1S, creates an instance of the definition, starts processing, and then if required stops it. A plug-in is written in C++ and is defined as a class whose sample implementation is as follows. class plugin { public: virtual bool setup(state *, command *); virtual void control(state *, command *); virtual bool cleanup(state *, command *); };

The class definition includes a setup function, a cleanup function, and a main routine which is executed in a control cycle of 5 ms in HRP-1S. Figure 3 shows the architecture of the plug-in manager. The plug-in manager consists of two threads. One thread gets the robot status (sensory information) from the Adapter, executes the main routine of each plug-in in order, and outputs the motor commands to the adapter. This thread is a real-time thread that runs in a control cycle of 5 ms. The other thread is a non-real-time thread that controls plug-ins according to the commands from the user interface and so on. These two threads share a instance table in which the plug-ins currently loaded are registered. As we mentioned before, ART-Linux enables the execution of real-time processes at the user level so that we can implement real-time threads in the same way as non-real-time ones. This feature of ART-Linux makes it possible for the real-time thread and the non-real-time thread to communicate with each other. The plug-in implemented as a class of C++ is compiled as a shared object. The shared object is dynamically loaded and put into the control loop by the plug-in manager when its function is needed. If a modification of a function in a plug-in is required, we can renew the plug-in without stopping the whole system by using the following sequence: 1. modify the source code of the plug-in; 2. remake the shared object from the modified code;


354


3. unload the plug-in;

In this section, we present some basic plug-ins.

center of gravity. By applying constraints to limit the motion of the humanoid robot as an inverted pendulum moving in a plane, we can obtain simple dynamics of the robot. It allows a separate controller design for the sagittal (x– z) and lateral (y–z) motions and simplifies walking pattern generation a great deal. This merit makes real-time walk generation possible.

4.1. HUMANOID Plug-in

4.4. STABILIZER Plug-in

The HUMANOID plug-in commands on and off operations of power and servo. It is also used to set servo gains of the motor drivers.

In order to keep the robot balanced, the STABILIZER plugin modifies the joint angle trajectories produced by the SEQPLAY plug-in according to the sensor data sent from the Honda I/O board via the shared memory. Even if the joint servo controllers work perfectly, it is impossible to make a perfect motion pattern because of errors subtending in a robot model and an environment model that are necessary to make dynamically consistent motion patterns. This requires a reflex controller to maintain the stability of HRP-1S. There are many reflex control methods for humanoid and biped robots (Azevedo, Poignet, and Espiau 2001; Hirai et al. 1998; Huang, Nakamura, and Inamura 2001; Kagami et al. 2000a; Park and Kwon 2001; Pratt and Pratt 1998;Yamaguchi et al. 1996). We have introduced the following three control subsystems to achieve stable motion of HRP-1S. We will describe the details of our reflex control method in another paper.

4. reload the new shared object.

4. Plug-ins

4.2. SEQPLAY Plug-in The SEQPLAY plug-in sets the command angles of all joints and passes it to the Honda I/O board through the I/O libraries. A motion pattern generated off-line by using a motion generator can be loaded by this plug-in. Note that the SEQPLAY plug-in just plays back a motion pattern and does not care whether the motion pattern is dynamically consistent or not. We have used some motion pattern generators. The first was developed by Hitachi and Waseda University in phase one of HRP (Nemoto et al. 2001). It can produce a biped locomotion pattern adaptive to a terrain including walking straight, turning, going up/down stairs. Its basic algorithm is proposed in Yamaguchi and Soga (1999). The second was developed by our group (Kajita, Matsumoto, and Saigo 2001). It is based on a dynamic modeling method called the “three-dimensional linear inverted pendulum mode” (3D-LIPM). We also made another plug-in “KWALK” using the same algorithm shown in the next subsection. The final motion pattern generator was developed by the University of Tokyo (Nakazawa et al. 2002). It is able to import a human dance motion, which is acquired from a motion capture system, into a humanoid robot. It is possible to use not only these motion pattern generators but also any type of generator that can produce a motion pattern which produces desired trajectories of all joints and ZMP (Vukobratovi´c and Stepanenko 1972) which also matches the data format of OpenHRP. 4.3. KWALK Plug-in Recently, some researchers have proposed a real-time generation method of humanoid walking motion (Yagi and Lumelsky 1999; Inoue et al. 2000b; Denk and Schmidt 2001; Hirose et al. 2001; Lim, Kaneshima, and Takanishi 2001; Nishiwaki et al. 2001, 2002; Sugihara, Nakamura, and Inoue 2002). The KWALK plug-in can also produce biped walking pattern in real time. It is based on a real-time motion generating algorithm based on 3D-LIPM (Kajita et al. 2001). When a biped robot is supporting its body on one leg, its dominant dynamics can be represented by a single inverted pendulum that connects the supporting foot and the robot’s

4.4.1. Body Inclination Control We adjust each foot’s position and orientation according to the difference between the desired body posture and the actual body posture, as if the robot was an inverted pendulum. Here the desired body posture is given by the motion generators. The actual body posture is estimated by a Kalman filter based on data from the body inclination sensors, which consist of gyroscopes and G-force sensors. During the single support phase, the orientation of the supporting foot is controlled around the desired ZMP given by the motion generator. It makes the rear section of the supporting foot lower when the robot’s body tips backward. During the double support phase, orientation and position of both feet are controlled around the desired ZMP within the convex encasing both feet. 4.4.2. ZMP Damping Control In order to reduce the error between the desired ZMP trajectory given by the motion generators and the actual ZMP calculated from contact force/torque information measured by sixaxis force/torque sensor integrated within the each foot of the robot, we adjust the horizontal position of the torso. When the actual ZMP is forward of the desired ZMP, the torso is accelerated forward. This control makes the actual ZMP as smooth as connecting a sky-hook damper to the torso. The motion of


Yokoi et al. / Humanoid Robot HRP-1S the torso is actually produced by the opposite movement of the support foot. 4.4.3. Foot Adjusting Control In order to move on an uneven terrain with bumps and dents, each foot should be adjusted to the inclination of the ground. The body inclination control mentioned above shifts the desired ZMP to an appropriate position that we call the m-ZMP. Even if the ground surface is inclined, the force/torque sensors in the feet detect changes in the ground reaction force. Using these changes, the actual ZMP is controlled to the m-ZMP by rotating each supporting foot. Note that the foot adjusting control is not necessary when HRP-1S walking on a flat terrain. 4.5. LOG Plug-in The Log plug-in stores all motion data including joint angles, body inclination, contact force, and so on.

5. Experiments In order to confirm the performance of the motion controller of HRP-1S, we performed several experiments. The absorption of the landing-impact force is very important for practical biped robots (Yamaguchi et al. 1996). HRP1S has impact absorption mechanisms in its feet quite similar to Honda P2 (Hirai et al. 1998). It is effective in reducing the transmission of impact forces, however it induces a rotational compliance between the ankle and the ground. In any motion generator, it is very hard to consider this compliance effect because of its high nonlinearity and the effect from the working environment in which unknown bumps and dents might exist. Figures 4–9 show scenes and data of HRP-1S walking six steps on a flat floor driven by a motion pattern generated by Hitachi/Waseda’s motion generator (Nemoto et al. 2001). The step length is 0.2 m per step and the step period is 0.8 s per step. Figure 4 (Extension 1) shows a sequence of photographs of HRP-1S while it was walking without the STABILIZER plug-in. The body inclinations and ground reaction forces are shown in Figure 5. Figure 6 shows the time histories of reference (dashed line) and actual (solid line) pitch angles of right and left hip joints (top), knee joints (middle), and ankle joints (bottom). In this experiment, the body inclined 0.1◦ around the roll axis and 3.4◦ around the pitch axis when HRP-1S started walking at 12 s, as shown in Figure 5. Although the actual joint angles followed the reference joint angles as shown, the body inclinations increased step by step. HRP-1S walked three steps, and it tipped back over after around 16 s. The body inclinations recovered after 16.4 s, because the body was hanging from a crane.

355

Even if the impact absorption mechanisms were installed in the feet, it was hard to reduce the impact caused by these huge unexpected inclinations. Consequently, the maximum ground reaction force was over 200 kgf which is 1.7 times the weight of HRP-1S. Figures 7–9 show the results while HRP-1S was walking with the STABILIZER plug-in (Extension 2). Note that we turn on only the body inclination control and the ZMP damping control of the STABILIZER plug-in when HRP-1S walks on a flat floor. The body inclined about 1◦ around a pitch axis when HRP-1S started walking. While it was walking, the body inclination around the roll axis was stable in the region of −0.4◦ to +0.6◦ and that motion around the pitch axis was a stable in the region of +0.2◦ to +1.2◦ . In Figure 9, the shapes of the time histories of the reference joint angles are slightly different from those in Figure 6, because the STABILIZER plug-in modifies them to maintain balance. Again, the actual joint angles followed the reference joint angles well. The maximum ground reaction force was about 131 kgf which is only 1.1 times of the weight of HRP-1S. Figures 10 (Extension 3) and 11 (Extension 4) show the experimental results of walking on a flat floor. In these experiments, we turn on only the body inclination control and the ZMP damping control of the STABILIZER plug-in. Figure 10 shows the experimental results of turning with a step length about 0.2 m per step, the step period 0.8 s per step, and the turning angle 15◦ per step. In this experiment, we used the KWALK plug-in and prepared the walk pattern off-line. Figure 11 shows the experimental results walking with 1 m on a straight line and turning left with 90◦ simultaneously with step period of 0.8 s per step. In this experiment, we used the KWALK plug-in and prepared the walk pattern on-line. Figures 12 (Extension 5) and 13 (Extension 6) show the experimental results of walking on an unknown rigid uneven floor. In these experiments, we turn on all control elements of the STABILIZER plug-in. Figure 12 shows the experimental results of walking on an uneven floor with a height of 20 mm or less and a gradient of 5% or less. HRP-1S can stably walk on the unknown and uneven terrain with a step length of 0.2 m per step and a step period of 0.8 s per step. Figure 13 shows the experimental results of walking on another unknown uneven terrain, which is made up of random paving of 400 mm2 plates with a height of 20 mm. HRP-1S can stably walk on the unknown and uneven terrain with a step length 0.2 m per step and a step period of 0.8 s per step. The humanoid robot HRP-1S is capable of various types of motion, not only biped walking. Figure 14 (Extension 7) shows gymnastic motion performed by HRP-1S. HRP-1S bends down and sideways, balances on one leg, and kicks. This motion is created off-line by using the dynamic filtering technique proposed in Kagami et al. (2000b).


356


Body Inclination [deg]

Fig. 4. Experimental results: HRP-1S walks without the STABILIZER plug-in.

5 4

Roll Pitch

2 0 -2 -4 -5 12

13

14

15 time [s]

16

17

18

200 Right Left

Fz [kgf]

150 100 50 0 -50 12

13

14

15 time [s]

16

17

18

Fig. 5. Body inclinations and ground reaction forces: HRP-1S walks without the STABILIZER plug-in.

As shown in these figures, we could successfully control humanoid robot HRP-1S using our own software.

6. Conclusions In this paper we have presented how we can control Honda’s humanoid robot without Honda’s control software. The humanoid robot HRP-1S was made by replacing the main control CPU and the software of HRP-1 that was produced by Honda. By introducing ART-Linux as an operating system, we developed the actual real-time controller for whole-body movements. The developed controller can be used in the sim-

ulation without any modifications. The motion controller is built in accordance with the plug-in architecture. We have developed some plug-ins including a real-time walking pattern generator (KWALK plug-in) and a reflex controller (STABILIZER plug-in) to maintain the dynamic balance of HRP-1S. The effectiveness of the proposed motion control system of HRP-1S was experimentally confirmed. Future work includes the development of a real-time whole-body motion generator and a more stable balance controller to adapt to rougher terrain. Implementation of a real-time collision checker is also imperative to avoid selfdestruction.



hipp [deg]

0 10 20 30 40

Right Left 12

13

14

15

16

17

18

13

14

15

16

17

18

13

14

15

16

17

18

knee [deg]

70 60

Right Left

50 40 30 20 12

anklep [deg]

10 0

Right Left

10 20 30 40 12

time [s]

Fig. 6. Reference and actual joint angles: HRP-1S walks without the STABILIZER plug-in.

Fig. 7. Experimental results: HRP-1S walks with the STABILIZER plug-in.


357


Body Inclination [deg]

358

5 4

Roll Pitch

2 0 -2 -4 -5 10

11

12

13 time [s]

14

15

16

200 Right Left

Fz [kgf]

150 100 50 0 -50 10

11

12

13 time [s]

14

15

16

Fig. 8. Body inclinations and ground reaction forces: HRP-1S walks with the STABILIZER plug-in.

0

Right Left

hipp [deg]

10 20 30 40

10

11

12

13

14

15

16

70

Right Left

knee [deg]

60 50 40 30 20 10

11

12

13

14

15

16

anklep [deg]

10

Right Left

0 10 20 30 40 10

11

12

13

14

15

time [s]

Fig. 9. Reference and actual joint angles: HRP-1S walks with the STABILIZER plug-in.


16


Fig. 10. Turning of HRP-1S.

Fig. 11. Simultaneous walking straight and turning left of HRP-1S.

Fig. 12. Straightforward walking of HRP-1S on unknown uneven terrain 1.

Fig. 13. Straightforward walking of HRP-1S on unknown uneven terrain 2.


359

360


Fig. 14. Gymnastic motion of HRP-1S.

Appendix: Index to Multimedia Extensions The multimedia extension page is found at http://www. ijrr.org. Table of Multimedia Extensions Extension Type Description 1

Video

2

Video

3 4

Video Video

5

Video

6

Video

7

Video

HRP-1S walks without STABILIZER plug-in HRP-1S walks with STABILIZER plug-in Left turn Simultaneous walking straight and turning left Straightforward walking on unknown uneven terrain 1 Straightforward walking on unknown uneven terrain 2 Gymnastic motion of HRP-1S

Acknowledgments This research was supported by the Humanoid Robotics Project of the Ministry of Economy, Trade and Industry of Japan. The authors thank Hajime Saito of AIST for his cooperation.

References Azevedo, C., Poignet, P., and Espiau, B. 2001. Moving horizon control for biped robots without reference trajectory. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 2762–2767. Denk, J., and Schmidt, G. 2001. Synthesis of a walking primitive database for a humanoid robot using optimal control techniques. Proceedings of the IEEE-RAS International Conference Humanoid Robots, Tokyo, Japan, pp. 319–326. Espiau, B., and Sardain, P. 2000. The anthropomorphic biped robot BIP2000. Proceedings of the IEEE International Conference on Robotics and Automation, San Francisco, CA, pp. 3997–4002.

Furuta, T., Okumura, Y., Tawara, T., and Kitano, H. 2001. ‘morph’: a small-size humanoid platform for behavior coordination research. Proceedings of the IEEE-RAS International Conference Humanoid Robots, Tokyo, Japan, pp. 165–171. Gienger, M., Löffler, K., and Pfeiffer, F. 2001. Towards the design of a biped jogging robot. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 4140–4145. Hirai, K. 1997. Current and future perspective of Honda humanoid robot. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Grenoble, France, pp. 500–508. Hirai, K., Hirose, M., Haikawa, Y., and Takenaka, T. 1998. The development of Honda humanoid robot. Proceedings of the IEEE International Conference on Robotics and Automation, Grenoble, France, pp. 1321–1326. Hirose, M., Haikawa, H., Takenaka, T., and Hirai, K. 2001. Development of humanoid robot ASIMO. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Workshop on Explorations towards Humanoid Robot Applications, Maui, Hawaii. Hirukawa, H., Kanehiro, F., and Yokoi, K. 2001. HRP develops OpenHRP. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Workshop on Explorations towards Humanoid Robot Applications, Maui, Hawaii. Huang, Q., Nakamura, Y., and Inamura, T. 2001. Humanoids walk with feedforward dynamic pattern and feedback sensory reflection. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 4220–4225. Inaba, M. 1993. Remote-brain robotics: interfacing AI with real world behaviors. Proceedings of the International Symposium on Robotics Research, Pittsburgh, PA. Inoue, H., Tachi, S., Tanie, K., Yokoi, K., Hirai, S., Hirukawa, H., Hirai, K., Nakayama, S., Sawada, K., Nishiyama, T., Miki, O., Itoko, T., Inaba, H., and Sudo, M. 2000a. HRP: Humanoid Robotics Project of METI. Proceedings of the IEEE-RAS International Conference on Humanoid Robots, Boston, MA. Inoue, K., Yoshida, H., Arai, T., and Mae, Y. 2000b. Mobile


Yokoi et al. / Humanoid Robot HRP-1S manipulation of humanoids: real-time control based on manipulability and stability. Proceedings of the IEEE International Conference on Robotics and Automation, San Francisco, CA, pp. 2217–2222. Inoue, H., Tachi, S., Nakamura, Y., Hirai, K., Ohyu, N., Hirai, S., Tanie, K., Yokoi, K., and Hirukawa, H. 2001. Overview of Humanoid Robotics Project of METI. Proceedings of the International Symposium on Robotics, Seoul, Korea, pp. 1478–1482. Ishida, T., and Kuroki, Y. 2001. Motion entertainment by a small humanoid robot based on Open-R. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Maui, Hawaii. Ishiwata, Y., and Matsui, T. 1998. Development of Linux which has advanced real-time processing function. Proceedings of the RSJ Annual Conference, Sapporo, Japan, pp. 355–356. Kajita, S., and Tani, K. 1991. Study of dynamic biped locomotion on rugged terrain: derivation and application of the linear inverted pendulum mode. Proceedings of the IEEE International Conference on Robotics and Automation, Sacramento, CA, pp. 1405–1411. Kajita, S., Matsumoto, O., and Saigo, M. 2001. Real-time 3D walking pattern generation for a biped robot with telescopic legs. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 2299– 2036. Kajita, S., Kanehiro, F., Kaneko, K.,Yokoi, K., and Hirukawa, H. 2001. The 3D linear inverted pendulum mode: a simple modeling for a biped walking pattern generation. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Maui, Hawaii, pp. 239–246. Kagami, S., Nishiwaki, K., Kitagawa, T., Sugihara, T., Inaba, M., and Inoue, H. 2000a. A fast generation method of a dynamically stable humanoid robot trajectory with enhanced ZMP constraint. Proceedings of the IEEE-RAS International Conference on Humanoid Robots, Boston, MA. Kagami, S., Kanehiro, F., Tamiya, Y., Inaba, M., and Inoue, H. 2000b. AutoBalancer: an online dynamic balance compensation scheme for humanoid robots. Robotics: The Algorithmic Perspective, Workshop on Algorithmic Foundations of Robotics. Kagami, S., Nishiwaki, K., Sugihara, T., Kuffner, J.J., Inaba, M., and Inoue, H. 2001. Design and implementation of software research platform for humanoid robotics: H6. Proceedings of the IEEE International Conference on Robotics and Automation, pp. 2431–2436. Kanehiro, F., Inaba, M., Inoue, H., and Hirai, S. 2000. Developmental realization of whole-body humanoid behaviors based on StateNet architecture containing error recovery functions. Proceedings of the IEEE-RAS International Conference on Humanoid Robots, Maui, Hawaii. Kanehiro, F., Miyata, N., Kajita, S., Fujiwara, K., Hirukawa, H., Nakamura, Y., Yamane, K., Kohara, I., Kawamura, Y.,

361

and Sankai, Y. 2001a. Virtual humanoid robot platform to develop controllers of real humanoid robots without porting. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Kanehiro, F., Inaba, M., Inoue, H., Hirukawa, H., and Hirai, S. 2001b. Developmental software environment that is applicable to small-size humanoid and life-size humanoids. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 4084–4089. Kanehiro, F., Fujiwara, K., Kajita, S., Yokoi, K., Kaneko, K., Hirukawa, H., Nakamura, Y., and Yamane, K. 2002. Open architecture humanoid robotics platform. Proceedings of the IEEE International Conference on Robotics and Automation, Washington, DC, pp. 24–30. Kato, J., Mori,Y., and Masuda, T. 1972. Pneumatically powered artificial legs walking automatically under various circumstances. Proceedings of the 4th International Symposium on External Control of Human Extremities, pp. 458–470. Konno, A., Kato, N., Shirata, S., Furuta, T., and Uchiyama, M. 2000. Development of a light-weight biped humanoid robot. Proceedings of the IEEE/RSJ Conference on Intelligent Robots and Systems, Takamatsu, Japan, pp. 1565– 1570. Lim, H. O., Kaneshima, Y., and Takanishi, A. 2001. Online walking pattern generation for biped humanoid robot with trunk. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 3111– 3116. Lorch, O., Seara, J. F., Strobl, K., Hanebeck, U. D., and Schmidt, G. 2002. Perception errors in vision guided walking: analysis, modeling, and filtering. Proceedings of the IEEE International Conference on Robotics and Automation, Washington DC, pp. 2048–2053. McGeer, T. 1990. Passive dynamic walking. International Journal of Robotics Research 9(2):62–82. Nagakubo, A., Kuniyoshi, Y., and Cheng, G. 2001. ETLHumanoid: a high-performance full body humanoid system for versatile actions. Proceedings of the IEE/RSJ International Conference on Intelligent Robots and Systems, Maui, Hawaii, pp. 1087–1092. Nakazawa, A., Nakaoka, S., Ikeuchi, K., and Yokoi, K. 2002. Imitating human dance motions through motion structure analysis. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, pp. 2539–2544. Nemoto,Y., Hattori, S., Hagiwara, T., Ichige,Y., Fujie, M., and Takanishi, A. 2001. Development of a basic motion control library for humanoid. Proceedings of the International Symposium on Robotics, Seoul, Korea, pp. 1489–1490. Nishiwaki, K., Sugihara, T., Kagami, S., Inaba, M., and Inoue, H. 2001. Online mixture and connection of basic motions for humanoid walking control by footprint specification. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 4110–4115.


362


Nishiwaki, K., Kagami, S., Kuniyoshi, Y., Inaba, M., and Inoue, H. 2002. Online generation of humanoid walking motion based on a fast generation method of motion pattern that follows desired ZMP. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, pp. 2684–2689. Park, J. H., and Kwon, O. 2001. Reflex control of biped robot locomotion on a slippery surface. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 4134–4139. Pratt, J., and Pratt, G. 1998. Intuitive control of a planar bipedal walking robot. Proceedings of the IEEE International Conference on Robotics and Automation, Leuven, Belgium, pp. 2014–2021. Sugihara, T., Nakamura, Y., and Inoue, H. 2001. Real-time humanoid motion generation through ZMP manipulation based on inverted pendulum control. Proceedings of the IEEE International Conference on Robotics and Automation, Seoul, Korea, pp. 1404–1409. Takanishi, A., Ishida, M., Yamazaki, Y., and Kato, I. 1985. The realization of dynamic walking robot WL-10RD. Proceedings of the International Conference on Advanced Robotics, pp. 459–466. Vukobratovi´c, M., and Stepanenko, J. 1972. On the stability of anthropomorphic systems. Mathematical Biosciences 15:1–37.

Wolff, K., and Nordin, P. 2001. Evolution of efficient gait with humanoids using visual feedback. Proceedings of the IEEE-RAS International Conference on Humanoid Robots, Tokyo, Japan, pp. 99–106. Yagi, M., and Lumelsky, V. 1999. Biped robot locomotion in scenes with unknown obstacles. Proceedings of the IEEE International Conference on Robotics and Automation, Detroit, MI, pp. 375–380. Yamaguchi, J., and Soga, E. 1999. Development of a bipedal humanoid robot: control method of whole body cooperative dynamic biped walking. Proceedings of the IEEE International Conference on Robotics and Automation, Detroit, MI, pp. 368–374. Yamaguchi, J., Kinoshita, N., Takanishi, A., and Kato, I. 1996. Development of a dynamic biped walking system for humanoid: development of a biped walking robot adapting to the humans’ living floor. Proceedings of the IEEE International Conference on Robotics and Automation, Nagoya, Japan, pp. 232–239. Yamasaki, F., Matsui, T., Miyashita, T., and Kitano, H. 2000. PINO the humanoid that walk. Proceedings of the IEEE-RAS International Conference on Humanoid Robots, Boston, MA. Zheng, Y. F. and Shen, J. 1990. Gait synthesis for the SD-2 biped robot to climb sloping surface. IEEE Transactions on Robotics and Automation 6(1):86–96.
