design approach of biologically-inspired musculoskeletal … · 2016-10-04 · yuto nakanishi,...

Journal Name

Regular Paper

Yuto Nakanishi1, ⋆, Shigeki Ohta2, Takuma Shirai3, Yuki Asano4, Toyotaka Kozuki5,Yuriko Kakehashi6, Hironori Mizoguchi7, Tomoko Kurotobi8, Yotaro Motegi9,Kazuhiro Sasabuchi10, Junichi Urata11, Kei Okada12, Ikuo Mizuuchi13 and MasayukiInabaAAAAAAAAA14

1-12,14Department of Mechano-Informatics, The University of Tokyo, JAPAN13Department of Mechanical systems Engineering, Tokyo University of Agriculture and Technology, JAPAN⋆ Corresponding author E-mail: [email protected]

Received D M; Accepted D M

DOI: 10.5772/chapter.doi

© 2012 FIRST AUTHOR; licensee InTech. This is an open access article distributed under the terms of the Creative

Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution,

and reproduction in any medium, provided the original work is properly cited.

Abstract In order to realize more natural and variousmotions like humans, humanlike musculoskeletaltendon-driven humanoids have been studied. Especially,it is very challenging to design musculoskeletal bodystructure which consists of complicated bones, redundantpowerful and flexible muscles, and large number ofdistributed sensors. In addition, it is very challengingto reveal humanlike intelligence to manage thesecomplicated musculoskeletal body structure. This papersums up life-sized musculoskeletal humanoids Kenta,Kotaro, Kenzoh and Kenshiro which we have developedso far, and describes key technologies to develop andcontrol these robots.

Keywords Musculo-Skeletal Humanoid, Bio-InspiredRobotics, Body Design Methodology, Behavior Intelligence

1. Introduction

What should we study on as next robot researches? In thiscase, a musculoskeletal robot represented by humanoidsis a good target when considering the frontier to which theevolution of intelligent robot research. A musculoskeletalhumanoid is the robot which has following characteristics:

• It has fullbody flexibility based on its spine structure.• It has redundant muscles which can drive complicated

joint structure.• It can easily embed elastic elements for force control to

muscle itself.• It can easily embed stiffness adjustment elements to

muscle itself.

On the other hand, it is difficult to realize intelligenceto manage fullbody motion and behavior due to itscomplexity of body structures and a large number of jointDOFs and actuators to control. Such a research can bevaluable in evolving the development of new research onhumanoid intelligence. From the the experimental studyon theoretical structure of intelligence, we researchershave to reveal human intelligence how to manage complexmusculoskeletal body with many bones, organs andmuscles which is covered by flexible skins like humanbody. Needless to say, we human beings not only moveour musculoskeletal body, but also live in social dailylife by more high-dimensional intellectual behaviors, suchas recognizing environments, making tools, using themdexterously and so on.

For musculoskeleltal humanoid researches, it is alsonecessary not only to reveal such a musculoskeleltal

1Yuto Nakanishi, Shigeki Ohta, Takuma Shirai, Yuki Asano, Toyotaka Kozuki, Yuriko Kakehashi, Hironori Mizoguchi,

Tomoko Kurotobi, Yotaro Motegi, Kazuhiro Sasabuchi, Junichi Urata, Kei Okada, Ikuo Mizuuchi and Masayuki Inaba:

Design Approach of Biologically-Inspired Musculoskeletal Humanoids

www.intechopen.com

International Journal of Advanced Robotic Systems

ARTICLE

1 Department of Mechano-Informatics, The University of Tokyo, Japan2 Department of Mechanical systems Engineering, Tokyo University of Agriculture and Technology, Japan* Corresponding author E-mail: [email protected]

Accepted 07 Dec 2012

DOI: 10.5772/55443

∂ 2013 Nakanishi et al.; licensee InTech. This is an open access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution, and reproduction in any medium, provided the original work is properly cited.

Yuto Nakanishi1,*, Shigeki Ohta1, Takuma Shirai1, Yuki Asano1, Toyotaka Kozuki1, Yuriko Kakehashi1, Hironori Mizoguchi1, Tomoko Kurotobi1, Yotaro Motegi1, Kazuhiro Sasabuchi1, Junichi Urata1, Kei Okada1, Ikuo Mizuuchi2 and Masayuki Inaba1

Design Approach of Biologically-InspiredMusculoskeletal HumanoidsInvited Paper

Int J Adv Robotic Sy, 2013, Vol. 10, 216:2013

Figure 1. Evolution of small humanoid fullbody action

intelligence but also to prepare how to use theseintelligence to our social life in near future. It is veryimportant to prepare basic intelligent robot research forhuman daily assistant behaviors by using conventionaltype humanoid with not musculoskeletal body. Topromote musculoskeleltal humanoid researches, it isalso important to build developing environments whichcan inherit results for common type of humanoids,such as HRP2[1], PR2[2] and so on. We expect thata musculoskeletal humanoid research becomes thescholarship to learn human musculoskeletal structure andfunction, the intelligent study how to manage its complexbody and the intelligent robot platform for life-assistivehumanoid research thanks to its body characteristics.And also we hope this musculoskeletal research areawill evolve intelligent robot technology as the researchfrontier. Based on this vision, this paper describeslife-sized tendon-driven fullbody musculoskeletalhumanoid researches we have studied so far, from thebody design methodology to realization of fullbodybehavior.

2. Importance of humanoid flexibility in fullbody motion

Whole body behaviors are the behaviors where the robotis easily subject to impact from whole body inertia andcontact with environments due to its fast and flexiblebody motion. We developed small humanoid robotswhose brain computer, which is wirelessly connectedto the robot body, is outside of its body. As shown inFigure 1, we conducted various behavior achievementsand evaluations by using these small robots[3]. Manybehaviors were achieved, such as locomotion with arms,biped walking, carrying objects, swinging based onsensing its body oscillation, iron bar motion, monkeybars motion, walking and kicking, rolling-over, using amobile base, and so on[4]. These small humanoids weighlight because its body consists of low stiffness and lightstructures and it has no computer. Thanks to light weightbody, a physical damage or wire cable disconnectionrarely occur even in conflict with the environment. Inthese robots, their rotary joints consist of multi-reductiongear and motor modules. The gear structure takes impacttorque of its joint. It has the advantage that the jointimpact becomes small and indestructible due to thecompact and lightweight body.

Figure 2. Example of a musculoskeletal humanoid’s motion using

its spine

Kenta Kotaro Kojiro Kenzoh

Height[cm] 123 130 135 90

Weight[kg] 19 20(23) 45 35

PWR[W] 423 515.5(587.5) 3869 3900

PWR/Weight[W/kg] 22.2 25.8(25.5) 86.0 111.4

Number of muscles 95 91(107) 117 80Neck 6 6 6 10Spine 40 16 24 26Arms 28 40 40 44Legs 20 26(42) 44 0Eyes 1 3 3 2

Number of joint DOFs 80 69 58 46Neck 15 9 9 9Spine 30 15 12 9Arms 20 26 18 9Legs 14 16 16 26Eyes 1 3 3 2

Table 1. Specifications of tendon robots developed

However, the rolling over is difficult by the robotwhich has a rigid trunk with broad shoulders. Rollingover is the action which is need to bend and twist thetrunk. And also the motion shaking the whole body, suchas climbing a ladder or exercising an iron bar, requirespliant trunk by flexible spine structures. In order touse a mobile base, a robot needs to grab handles withtwo hands, dynamically deal with closed-link structurefrom arms to legs and strongly kick the ground by itslegs. This motion also requires whole body flexibility.In sitting in a chair and picking a object on the ground,redundant joint degree of freedoms are useful as shown inFigure 2. In whole body behaviors, the robot body needsflexibility which enables to absorb impacts by contactwith environments while shifting of the weight of thebody, and also mechanism which enables to achieve fastand agile motion. From this standpoints, it is effective toadopt musculoskeletal body structure, which can easilyimplement a spine structure and flexible joint structures.We expect that these musculoskeletal robot can evolve therobotics researches to new stages, such as realization ofmore humanlike dynamic fullbody behaviors.

3. Body design concept of musculoskeletal humanoids

This section describes the design concept of three life-sizedmusculoskeletal humanoids which we have developedso far. Figure 3 shows appearances, joint arrangementsand muscle arrangements of these robots. Table 1 showsheight, weight and number of muscles of each humanoid.

2 Int J Adv Robotic Sy, 2013, Vol. 10, 216:2013 www.intechopen.com

•Height : 123cm

•Weight: 19kg

•Joint DOFs: 73

•Muscle Num:94

•Height : 130cm

•Weight: 20kg

•Joint DOFs: 63

•Muscle Num: 108

•Height : 135cm

•Weight: 45kg

•Joint DOFs: 58

•Muscle Num: 109

Kenta(2000~) Kotaro(2005~) Kojiro(2008~)

Spherical

joint(3Dofs)

Rotational

joint(1Dof)

Tendon path

Joint arrangements

Tendon arrangements

Figure 3. Joints and tendon arrangements of musculoskeletal

humanoids developed.

3.1. fully tendon-driven musculoskeletal humanoid with spine:Kenta

We developed "Kenta"[5], which was developed to aimimplementation of more humanlike body structure basedon the former small humanoid "Cla" with spine joints[6].The big characteristics of Kenta is a complicated trunkstructure with super multi-DOFs vartebrae, which consistsof 10 articulated 3 DOFs joint (i.e. 30 DOFs) anda large number of redundant 40 muscles to drive itsvartebrae, as shown in Figure 4. Kenta’s vartebranot a simple ball-socket joint structure, but a complexstructure. It imitates a human vartebrae and has alock mechanism, which corresponds to human vartebraprojection for prevention of excessive movement. Thismechanism prevents its spine structure from bending atthe position of one vartebra. These joint structure withcomplicated shape was newly made of molded plasticresin by Selective Laser Sintering (SLS) method, which is akind of a rapid prototyping molding, not of conventionalengineering plastic or metal by cutting work. We haveimplemented humanlike complex bone structures of our

Kenta’s spine structure (back view)Motor array unit

Kenta

Tendon path

Motor

Figure 4. Accumulation of fixed muscles

musculoskeletal humanoids since Kenta until Kenshiro(insection5) by this SLS rapid prototyping.

In addition, Kenta has a neck structure with 5 articulatedcervical vertebrae driven by 6 tendons, and also its alljoints of its arms and legs are driven by tendons. This robotis the first life-sized fully tendon-driven musculoskeletalhumanoid which was developed at our lablatory. On theother hand, compared to its complex torso, its joint DOFsarrangement and tendon arrangement are simplisticallyimplemented. Kenta has a lot of room for improvementto achieve more humanlike structures.

3.2. Reinforceable musculoskeletal humanoid: Kotaro

We developed "Kotaro"[7], which was improved in thejoint DOFs arrangement and structures of its whole body,compared with previous humanoid Kenta. The basicactuation system of Kotaro is the same as one of Kenta,which adopts tendon driven system with winding pulleyrotated by DC 4.5W motor. Inspite of less numberof spine joints (i.e. 5 articulated spherical joints) thanKenta’s spine joints, Kotaro has a compound shoulderstructures consisting of a collarbone and a scapula. Asa result, it can humanlike upperbody natural movementthanks to humanlike arm joint DOFs arrangement, suchas its shoulder[8]. In addition, Kotaro can easily addand modify its tendon arrangements by its slim bonestructure and muscle modules with smooth round shapesas shown in Figure 5. Thanks to this design approach,we can test and search various tendon arranemgents ofKotaro after the robot actual implementation. We call thisconcpet "reinforceable musculoskeletal humaonid", whichwas proposed in [9]. This concpet is important in orderto reduce(i.e. simplify) number of muscles for actualimplementation to realize humanlike musculoskeletalhumanoids. And also all bone parts of Kotaro aremade of nilon resin by rapid prototyping method(SLS),which is used in Kenta’s spine structure. Kotaro enablesto implement more muscle actuators and more similartendon arrangements than Kenta, because we can freelydesign more complex bone shapes thanks to using rapidprototyping.

In the legs, Kotaro has as almost twice muscles as Kentaand also Kotaro adopts a DC 20[W] motor1 as an actuatorof antigravity muscles, which needs more power than

1 Kenta adopts a DC 4.5[W] motor




www.intechopen.com

FrontSide

Sensor board

Pulley

DC Motor

Beam for strain gauge

Deal Drawing

Wire

Meshed Hollow Skeleton

Section of frame

Motor Driver

Storing boards

Endoskeleton Structure

Muscle Unit

Thigh bone

Pelvis bone

Attachment pivot

Kotaro

Figure 5. Reinforceable musculo-skeletal humanoid Kotaro

others. Thanks to this improvement of actuation power,Kotaro can somehow keep standing[10], which Kentanever can do. However, it is difficult to achieve behaviorswhich needs enough legs power, such as standing up,squwatting, walking and so on. And also in the arms,Kotaro has a larger joint movable range than Kenta thanksto compound shoulder mechanism, however, it cannot putits hand up due to power lacking.

3.3. Highpower musculoskeletal humaonid: Kojiro

We developed "Kojiro"[11], which has high-poweractuation system, in order to solve lacking of joint powerin the previous musculoskeleltal humanoid Kenta andKotaro. The actuation system can instantly generate highpower to the limitation of AC 40[W] motor based onthermal control of motor’s inner coil[12]. Kojiro adoptsthe new design method[13], which enables to embedmany actuators and achieve high mechanical strength asshown in Figure 6. Thanks to this design approach, Kojirolegs can embed 44 actuators2, which is enough strong toconduct fullbody motion such as walking. Figure 7 showsKojiro’s complex leg structure with many actuators andcontrol boards.

And also Kojiro adopts newly spherical chest shoulderjoint mechanism[14], which enables to move smoothscapula with less friction than a conventional scapulastructure. In addition, Kojiro’s wrist joint can absorbimpact shock when it hits a nail, by embedding avariable joint stiffness mechanism. This mechanism isimplemented by putting nonlinear spring element inantagonistic tendons around its wrist.

3.4. Musculoskeletal humanoid driven by mechanicallycompliant tendons: Kenzoh

In order to let musculoskeletal humanoids do moredynamic motions with some complex contact toenvironments like sports, we developed musculoskeletalhumanoid "Kenzoh" (Figure 8) with mechanicallycompliant joint.

2 Kotaro legs has 20 actuators.

Muscle1DOF

3DOF

1.Decision of DOF 2.Decision of parts 3.Shape design

Outer shape

4.Hollowing out the assenmbly areas

Assembly areas for parts

After hollowing out

5.Building the tunnels for the cables

Cables TunnelsTu

Final lower leg bone structure

Kojiro

Figure 6. Design method of Kojiro’s skeleton(a shank bone)

Joint ArrangementJoint Arrangement Cable ArrangementCable Arrangement Parts ArrangementParts Arrangement Lowerbody OverviewLowerbody Overview

3DOF

1DOF

3DOF

1DOF

786[mm]

Actuator unit

DC/DC convertor USB hub Sensor board Motor driver

Robot Links

Thigh bone

Foot(Upper)

Crus bone

Foot(Lower)

Toe

6axes force sensor

Figure 7. The lower body arrangement of Kojiro

Figure 8. Musculo-skeletal humanoid "Kenzoh"


Spring

Wire

Pulley

Low stiffness High stiffness

Figure 9. Mechanism of non-linear spring tension unit(NST)

40W Actuator Unit with NST

90W Actuator Unit with NST

Actuator Elements of Chest bone

1.Motor

2.Spring

3.Guide Pulley

4.Wire(Tendon)

5.Tension Sensor

(Loadcell)

6.Pulley1

2

6

1

2

3

3

44

5

5

6

Figure 10. Integration NST unit with actuator muscle system

This robot can absorb impact shocks by using itsjoint softness based on mechanical stiffness adjustmentmuscle modules, which is called nonlinear spring tensionunit(NST) like in Figure 9. If robot’s joint is adjusted aslow stiffness by NST, it can be adapted against suddenlycollision with environments without no joint feedbackcontrol, because low joint stiffness is mechanicallyachieved by NST units. Of course, low joint stiffnesscan be achieved by joint feedback control by joint torquesensor[15](or tension sensor in case of tendon robots).However, it is difficult to achieve enough fast response toimpulse force, and also it does not work by some failure ofa torque sensor, which often is fragile. Therefore, in orderto develop adaptive humanoids with environment, robot’sjoints should be designed to control its own joint stiffnessmechanically.

One of most important Kenzoh characteristics is thatKenzoh’s all muscles have NST units. In order to embedmore than 80 number of NST units within Kenzoh’slimited body space, Kenzoh adopts two types of NST unitdeveloped by ourselves. One is the integration type withactuator unit parts as shown in Figure 10. Advantageof this NST unit is to utilize effectively motor shaftspace as spring and guided pulley of NST units, whichoften becomes dead space in previous musculoskeletalhumanoids we developed. In Kenzoh, the integration NSTunits are mainly used in a chest bone, bladebones andupperarms, which have enough spaces. The other NSTis the rotary NST unit as shown in Figure 11. Advantageof this NST unit is to put it anywhere through its tendonpath, which enables to widen freedom of robot design.It is very useful at designing a complicated structures ofKenzoh which has many elements. And also this rotaryNST unit can achieve higher stiffness than other NSTmechanisms[16]. In Kenzoh, the rotary NST units aremainly used in a spine vertebrae, which have less spaces.

Spring

Rotation axisTop VIew

Wire(Tendon of muscle)

Figure 11. Rotary NST unit mechanism

Motion generatorMotion generator

Motion controller

Robot bodyMotion system

Outer sensorOuter sensor

Inner sensorInner sensor

Middle&Lower layer

Actuator

Inner sensor

Actuator

Inner sensor

Actuator

Inner sensor

Actuator

Inner sensor

Body controller

Enviro

nm

en

t

Robot model

Information processing

Reference of

operational object

Modified

joint order

Upper layer

Original motion joint order

+-

++

Rmodθ

nowR inS

outS

orgθ

Figure 12. Upper layer of motion controller for musculoskeletal

humanoids.

Motion controller

Robot body

Inner sensorInner sensor

Middle layer

En

viro

nm

en

t

Information processingInformation processing

+

-

+

+

inS

Joint-Muscle Converter

(feedforward controller)

Muscle controllerMuscle controller

Muscle controllerMuscle controller

Muscle group A

Joint Feedback controllerJoint Feedback controller

Joint controller AActuator

Inner sensor

Actuator

Inner sensor

Actuator

Inner sensor

Actuator

Inner sensor

Joint controller B

Joint controller C

Lower layerUpper layer

Robot modelRobot model

Middle layer

Modified

joint order modθ ffl

fbl

nowjC

jCJoint control

parameter

modl

Muscle group B

Muscle group C

Figure 13. Middle layer of motion controller for musculoskeletal

humanoids.

4. Achievement of musculoskeletal humanoids behavior

Researches of achievement of musculoskeletal robotsbehavior have been studied based on mainly twomethodologies. The one is by controlling its joints basedon goal joint angle trajectory calculated like conventionalaxis-driven robots. The other is by learning and searchinghow to move its own complicated body itself. In thecase of controlling joint angle, tendon length trajectoryis converted from aimed joint angle trajectory by usingconversion methodology as we discuss later, and eachtendon length is controlled based on calculated trajectorydata as shown in Figure 12. However, this method hasthe problem that precise joint control is difficult due toelongation of tendon wires and the error of tendon paths inthe geometric model which is used in ’joint angle - tendonlength’ conversion process. Therefore, in the case whichneeds accuracy of joint angles or end effector position,sensor feedback control is necessary based on sensorswhich can detect its joint angle or hand positions as shownin Figure 13.




www.intechopen.com

Feedforward control

1

Feedback control for balancing

2

Feedback control for balancing

2

Muscle reflex based on tension

3

Muscle reflex based on tension

3

Figure 14. Proposed balancing approach for musculoskeletal

humanoid

From these background, learning and searching methodseems to be more fit for achievement of musculoskeletalhumanoids behaviors rather than conventional precisejoint control method, however, there are no establishedmethod in these robot research field, which is currentlyat the stage of many researchers exploring the methodthrough their trial and error approach as following:Creeping motion acquisition based on modificationthrough initial motion through genetic algorithmsearching[17, 18], swinging behavior achievement basedon the sensor-guided methodology by genetic algorithmbased modification of the neural network whose inputis sensor value, such as gyroscope and accelerometerand output is actuator state[19], achievement of steppingmotion based on sensor reflex[20] and auto-tuningmethods of its reflex control parameter in the realworld[21], pedaling motion acquisition by modification offorce control parameter through trial and error in the realworld [22]. Narioka et al. realized creeping behavior bypneumatic tendon-driven musculoskeletal robot based onon-line optimization of CPG parameters[23].

In addition, by constructing hierarchical sensor feedbackstructure which consists of combination of spinal cordreflex and voluntary movement like human controlstructure(Figure 14, robust control method against theerror between the control model and the real robot bodyis proposed [24, 25], such as keeping balance as shown inFigure 15 and dynamic ball catching. Further, in order toachieve these musculoskeletal humanoid behaviors, it isimportant to maintain a continuous long-term operationby using its complex physical body. The following arekey technologies in supporting the achievement of thesemusculoskeletal humanoid movement.

4.1. Joint posture estimation methodology

It is a difficult task to obtain joint posture of amusculoskeleltal humanoid which has complicated bonestructures like human body. Mainly two methods areproposed to detect musculoskeleltal robot’s joint posture.The one is joint posture estimation from muscle-tendoninformation, and the other is to measure directly jointposture by newly developing bone posture detectionsensors.

In tendon-driven robots, one joint posture can hasredundant actuators status, which means that actuatorstatus can be underspecified from one joint posture status.Length set of tendons in a joint angle is easily obtainedby calculating geometric paths (i.e. backward kinematics),however, it is difficult to obtain joint angle from length

Figure 15. Demonstration of Kojiro bending/extending motion

using its hip and spine joints with proposed balancer

set of tendons (i.e. forward kinematics). The relationship

between joint angle�θ and tendon length�l is described

as jacobi matrix Jm(δ�l = Jmδ�θ). Actually, joint angleestimation methods are proposed[19] by using repeatedcomputation based on minimally changing joint angle

δ�θ = J♯mδ�l, where J

♯m is pseudo inverse matrix of Jm , and

using neural network.

In addition, by focusing on nonlinear relationship betweenthis joint angle and tendon length as shown in Figure16, joint posture estimation method by using only tendonvelocity component information is proposed[26]. Thismethod enables to estimate joint posture from tendonlength information while humanoid moving without thecalibration task to detect base posture. Figure 17 is theflowchart of this estimation method. These estimationmethods from tendon length information are suitable jointangle acquisition approach for complicated joint, such asspherical joint structures in human body. However, theyhave the problem that it is difficult to estimate precise jointangle due to tendon elongation and fracture.

On the other hand, we have also developed sensors whichdetect directly posture of a spherical joint, such as thesensor by using magnetic field change[27] and the sensorby tracking spherical surfaces pattern by using tiny camerawhich is used as cellular phone camera as shown in Figure18[28].

4.2. Wearable suits based on muscle actuator units

In order to reinforce muscle and re-configure musclearrangement after developing humanoid, reinforceable


Joint angle Muscle length L

Link Pulley

L

dLd

= const

Samples of muscle

"relative" length

Joint-Length Space

Impossible to identify pattern!

l

Figure 16. Joint pose estimation in the case of a simple joint

structure

Base joint posture:P(t0)

DataBase:f[k]

l(t0) l(tnow)Sample Data: Lr

l(t1)

P0 P1 PN

L0 L1 LN

Pattern matching

for decision of P(t0) Relative->Absolute

Now relative muscle length

l(tnow)

Now absolute muscle length

L(tnow)

t = t0 t = t1 t = tnow

Real robot

Relative muscle length (encoder)

Posture:P Muscle length:La

Map f :posture to muscle length

Quantization of Map f

Now joint posture: P(tnow)

Current posture estimation

Figure 17. Flowchart of the proposed joint estimation algorithm

Figure 18. Sensor for spherical joint using a small camera

musculoskeletal humanoid Kotaro was designed. Forachievement of this reinforceable muscle concept,muscle actuator is developed as a unit which can beindependently actuated[9, 29]. This muscle actuator unithas tension sensor built-in, and it can control tendontension based on its sensor on 1[kHz] cycle. Tendontension control with constant goal tension force has beenused to teach and playback musculoskeletal humanoids’postures and motions[30].

By using this muscle actuator unit, we developed thewearable manipulation suit[31] as shown in Figure 19.This suit is basically a lacrosse protector, where 24muscle actuator units are arranged across spine andarm joints. By controlling all units based on goaltension value, each muscle unit’s tendon length is

Figure 19. The wearable device using muscle actuator modules.

0

2

4

6

8

10

12

14

16

0 5 10 15 20

40

42

44

46

48

50

52

Ten

sio

n [

kg

f]

Len

gth

[m

m]

Time[s]

Tension

Length

0

2

4

6

8

10

0 5 10 15 20

-100

-90

-80

-70

-60

-50

Ten

sio

n [

kg

f]

Len

gth

[m

m]

Time[s]

Tension

Length

Figure 20. Tension of robot and wearable device during table

cleaning motion.

Figure 21. Motion control by a wearable device using muscle

actuator modules.

automatically changed to fit according to wearer’s posture.By estimating wearer’s posture from each muscle unittendon length information and sending wearer’s postureto the musculoskeletal humanoid as goal posture, thewearer can easily manipulate and navigate the robot.Vice versa, by modifying goal tension value of muscleunits on the wearable suit according to the tensionsensor value on the navigated robot, the wearer canfeel robot’s reaction force from environment. Figure20 shows muscle tension transition in the navigatedrobot(the left) and the wearer(the right) sides. By thisbilateral-navigation-control system, the wearer can teachnot only robot motion and also reaction force in its hand,which is very useful to achieve behaviors with contact toenvironments, such as wiping table (Figure 21) and so on.

4.3. Wholebody distributed motor-sensor control system

In super multi degree of freedom musculoskeleltalhumanoid with over 100 actuators and 500 sensors,it is necessary to deal with large number of thesesensor-motor information efficiently. Base concept ofinformation communication structure is the distributedmotor-sensor control system, where distributed smallprocessors individually control local actuators and collectlocal sensors on real-time 1[kHz] cycle, and central mainprocessor communicates with each distributed processoron slower 100[Hz] cycle[32].




www.intechopen.com

Figure 22. Example of warning on robot information viewer.

Figure 23. Communication configuration

Firstly, we developed the small motor driver board(36[mm] x 46[mm]) which can drive 4 axes DC4[W] motorsand continuously generate 5[A] and each motor driverscommunicate with main central processor through USBcommunication[33]. In these large distributed controlsystem, it is important to visualize whole sensor-motorcontrol status to detect disordered parts and preventbreakdown. Actually we developed visualization system[34, 35] as shown in Figure 22. And also, it is importantweight power ratio in the robot actuating system. FromKojiro robot, we have adopted high-power motor driver,which instantly drives motor with over its rating powerbased on motor inner coil’s thermal estimation[12].Thanks to this high-power driving system, Kojiro cangenerate instantly 10 times greater power than olderrobot Kotaro. Actually we developed new motor driverH8-18 based on this thermal control methodology. H8-18is equipped with microcomputer H8S/2215 of RenesusTechnologies Ltd. Co.Afs make and FPGA Cyclone IIIof Altera Inc.Afs make. With H8S/2215 and FPGAbeing interconnected by a bus, H8S/2215 communicateswith the host computer via USB (12 Mbps) and FPGAperforms communications on RS485 with LTC2854 thesame way as FPGA-01D. Figure 23 shows H8-18 andcommunication diagram and other systems that are usedfor driving. Assuming that a maximum of four boardsshould be connected per H8-18, communication speeds of

Figure 24. Musculoskeletal humanoid Kenshiro

RS485 have been set to 3 MHz and data are exchangedthrough asynchronous serial communication. This newboard is adopted by our robots Kenzoh(in section 3) andKenshiro(in section 5).

5. Challenge to realize a musculoskeletal humanoid as ahuman body simulator

As we mentioned above, we have developed andcontrolled complicated musculoskeletal humanoids,which have about 70 DOFs joints comprising a spine, ascapula and collarbone joint, and over 100 redundanttendon actuators.

However, these robots are designed by changing andsimplifying the tendon arrangements and bone shapes ofthese robots as compared to human body, due to limitationof implementation technology, especially bone materialsand actuator systems. They are also inadequate as a detailhuman body simulator in the real world.

Therefore, in order to solve these problems, we are nownewly developing a musculoskeletal humanoid, whichis named "Kenshiro"(see Figure 24). Kenshiro imitateshuman body with great accuracy, such as especiallyfollowing points:

• Weight ratios and sizes of each bones/limbs• Muscle arrangements and joint arrangements• Muscle output power

The goal of this robot is the achievement of both 1)veryhumanlike joint and muscle arrangements and 2)actuatorperformances for fullbody motion.

5.1. Human bones imitation in weight and sizes

Figure 25 shows kenshiro specification. Kenshiro isdesigned so that its height becomes shortest as long asit can contains 100 new actuators. Finally, the heightof Kenshiro will be in the order of 158[cm] and itweight approximately 50[kg]. This corresponds to almostJapanese 12 aged boy’s height and weights according tothe statistical data by Japan government 3. Each bonelink’s size and weight are determined by proportionallyconverting the statistical data(size-JPN 2004-2006) 4 of

3 http://www.mext.go.jp/bmenu/toukei/chousa05/hoken/kekka/kdetail/1303380.htm

4 http://www.meti.go.jp/press/20071001007/20071001007.html


Joint DOFs arrangement

3 DOFs spherical joint

2 DOFs oval spherical joint

1 DOF rotational joint

2 DOFs universal joint

Bone arrangement(Front , Back) Bone and actuator arrangement(Front )Bone, actuator and muscle tendons

arrangement(Front , Back)

158cm, 50kg

Figure 25. Outline of Kenshiro as a detail human simulator robot

Japanese young man with 170cm tall, which is publishedby NEDO 5, to the goal height 158cm.

5.2. Human joint arrangement imitation

Kenshiro has 64 joint DOFs except hands DOFs, whichconsists of a neck 13 DOFs, each arm 13 DOFs, each leg7 DOFs and spines 11 DOFs, as shown in Figure 25. Inthese joint structures, especially its spines and knee joint isdifferent form conventional musculoskeletal humanoids’ones.

Conventional spine structures before Kenshiro aredesigned as from 3 to 10 series articular joint structures,of which each joint is implemented as 3-DOFs sphericaljoint mechanism. However, in actual human body, thespine joint movable range has aeolotropies. For example,human 12 series thoracic vertebrae can rotate ±20[deg]in roll axis, from +45[deg] to -40[deg] in pitch axis and±35[deg] in yaw axis, and also human 5 series lumbarvertebrae can rotate ±20[deg] in roll axis, from +60[deg]to -20[deg] in pitch axis and ±5[deg] in yaw axis[36].According to this anatomical data, we can regard thathuman lumbar vertebrae seldom contribute yaw axisrotation. Therefore, in Kenshiro’s spine structure, lumbarvertebrae is designed as a 5 series articular joint structure,of which each joint is 2 DOFs mechanisms which canrotate in only roll and pitch axis and cannot rotate in yawaxis, as shown in Figure 26. A chest is designed as a solidconstruction with no movable parts. Inside of chest part,30 actuators to drive arms, lumbar vertebrae(i.e. spine)and cervical vertebrae(i.e. neck) must be embedded. Ifthoracic vertebrae moved, we would have had to solve theproblems of interferences between chest bones(i.e. costalbones) and 30 actuator systems within chest part. Thisproblem is very difficult to solve, therefore, at present,we designed Kenshiro’s chest as one solid part by fixing

5 New Energy and Industrial Technology Development Organization

Figure 26. Kenshiro spine joint arrangement

each thoracic vertebrae joints. However, Kenshiro’s chestconsists of 12 thoracic vertebrae similar to actual humanchest. We aim to realize movable chest part in near futureby unfixing each thoracic vertebrae.

And also Kenshiro has one DOFs rotational joint in yawaxis, between the bottom of its solid chest part and thetop of lumbar vertebrae. With these approximations,Kenshiro’s spine will achieve approximately the samerange of movements of humans.

Conventional knee joints before Kenshiro are designedas simple rotational joints or rolling joints, however, thehuman knee joint mechanism is much more complicated.Etoundi et al. developed a knee joint which has parallellinkage structures as human cruciate ligaments[37].Kenshiro’s knee has also parallel linkage structure and, inaddition, a floating bone as a patella. And also Kenshiro’sknee structure enables to achieve a screw home movement,




www.intechopen.com

Figure 27. Kenshiro knee mechanism

which is the human knee’s typical movement. Humanknee can be twisted when its knee is bent, and also itcannot be twisted when it is extended. It means humanknee’s movable range around twisting axis depends onits bending angle. Figure 27 shows Kenshiro’s kneemechanism.

5.3. Human muscle arrangement imitation

The muscle arrangement of Kenshiro imitates humanmain muscle arrangements as shown in Figure 25.Especially, geometric positions of muscle insertionpoints(i.e. beginpoints and endpoints) in bones areprecisely imitated based on real human anatomical bonedata model. And also we design as similar bone shapeto human’s as possible, because outer shapes of boneshave influence on interference between bone surface andmuscle tendons.

Finally, Kenshiro has more muscles, that are also betterarranged than previous musculoskeletal humaniod robots.Actually, it has 102 muscle actuators except head, neckand hands, and 160 muscles, which consist of 50 musclesin its legs, 76 muscles in its trunk part, 12 muscles inits shoulder and neck parts and 22 muscles in arms. InKenshiro, the number of DOF is more than the number ofactuators. This is because that we adopt a planar musclemechanism as Kenshiro’s muscle in trunk body, such asits abdominal muscles, latissimus dorsi and so on. Planarmuscle mechanism enables to move several muscles byonly one actuator by using moving pulley.

5.4. Muscle power improvement for fullbody motion

Except 8 muscles for wrist drive, all 94 muscle actuatorsconsist of Maxon 100[W] AC motors, therefore, totalrated power output of Kenshiro becomes about 10000[W]and rated power output per unit mass becomes about200[W/kg]. Actually, Kenshiro total rated power outputcorresponds to as 5 times as Kojiro’s one. Kojiro isthe previous musculoskeletal humanoid we developed in2008. It can somewhat walk, stand up, and do somefullbody motions[21, 24]. And Kenshiro rated powerper unit mass corresponds to as 3 times as Kojiro’sone. Actually Figure 28 shows the torque generated atthe hip joint in Kojiro(blue) and Kenshiro(green). And

Figure 28. Generating torque in Kenshiro and Kojiro hip pitch joint

Knee and pelvis bone by aluminum cutting work

Chest ’s costal bone:

aluminum casting

Chest ’s thoracic vartebrae :

aluminum cutting work

Collar bone made by Metal SLS

Collar bone made of CFRP

Figure 29. Bone structures of Kenshiro

also, Kenshiro’s knee joint can generate continuous torque116[Nm] and noload angular velocity is 19.6[rpm]. Thisspecification is comparable to general fullbody axis-drivenhumanoid HRP2 [1, 38] which can walk.

Thanks to this large improvement of power output, weaim to make Kenshiro do fullbody motions, which requirestrong motor intensity, such as walking with humanlikespine motion, standing up motion from sitting on a chair.It is very important in order to make Kenshiro the actualhuman body simulator to evaluate each muscle loadduring walking.

5.5. Complicated and strong bone design imitating humanlikebone shapes

As mentioned above, we have to design preciselyhumanlike outer bone shapes, in order to imitate humanmuscle arrangement in not only insertion points but alsotendon paths, because there are interferences betweenmuscle tendons and outer surface of bones and thisinterference effects depend on clearly shapes of outer bone.In the musculoskeletal humanoids we developed so far, webuild bones based on rapid prototyping processes, suchas SLA and SLS and so on, to realize humanoid boneswith humanlike complex curved surface shapes. However,these bone materials which can be manufactured by these




www.intechopen.com

rapid prototyping, such as ABS plastic resin and nylonresin, is much more fragile than metal materials, such asaluminum alloy. As a result, we have to change boneshapes to improve bone’s strength. Therefore, as long aswe adopt these fragile materials as humanoid structure, itis difficult to precisely imitate human bone shapes, whichhas many small outshoot parts and thin parts, such as apelvis bone, vertebrae, a costal bone, a scapula bone andso on.

Therefore, almost all bone structures of Kenshiro are madeof strong materials, such as aluminum alloy, steel alloyand CFRP and so on. Thanks to these materials, evenhumanlike small outshoot parts and thin parts can bebuilt without any fragility. On the other hand, thesemetal materials are more difficult to product than thematerials by rapid prototyping manufacture, especially inthe complicated bone structure with humanlike curvedsurfaces. It means that the bones cost more. InKenshiro, in order to reduce these bone design cost,we select manufacturing methods depending on boneshapes. Figure 29 shows Kenshiro’s bone structures.For example, costal bones are produced by aluminumcasting based on lost-wax process, because costal bonesconsist of about 30 parts and also, each part shapes arequite complicated with curved surfaces. Pelvis bone isproduced by aluminum cutting work, because pelvis boneconsists of only two parts(i.e. left and right part) and thecomplex shape consists of convex curved surfaces, whichcan somewhat easily be worked by cutting. Vertebra boneis also produced by aluminum cutting work by separatingvertebra bone into two parts(i.e. a main joint part anda stylocerite part), of which each part can be simplifiedand easily worked by cutting. Two kind of collar bonesare test-produced. The one is made of CFRP material,which enables to make a collar bone a long thin structure,because CFRP is more stiff than other materials, such asaluminum and steel. The other is made by metal SelectiveLaser Sintering, which enables to make a collar bone shapevery complicated as shown in Figure 29.

6. Conclusion

Robotics field is a frontier of autonomous systemwhich integrates evolution of science, engineering andtechnology. Many young researchers have becomeenthusiastic in this field with ambitious dreams. Especiallyin this paper, we introduced studies on life-sizedmusculoskeletal humanoids which has humanlike reallycomplicated body, as a research theme where researchershave to keep pursuing new intelligent robotics frontier.By setting these humanlike musculoskeletal humanoidresearches as one of big theme of our laboratory, wehave to keep thinking about evolution of intelligentrobotics that is how to design these complicated bodystructure and how to manage these complicated body. Itmay take a long time to realize a great musculoskeletalhumanoid, however, we feel that humanlike flexibility andredundancy are necessary in order to achieve a partnerrobot which support our life with us in future. Thispaper concludes that these musculoskeletal humanoidshave enough potential and possibility to become a future

humanoid platform in both daily assistive robotics fieldsand intelligence robotics fields.

7. References

[1] Kenji KANEKO, Fumio KANEHIRO, Shuuji KAJITA,Kazuhiko YOKOYAMA, Kazuhiko AKACHI,Toshikazu KAWASAKI, Shigehiko OTA, andTakakatsu ISOZUMI. Design of prototype humanoidrobotics platform for hrp. In Proceedings of the2002 IEEE/RSJ International Conference on IntelligentRobots and Systems, pages 2431–2436, Lausanne,Switzerland, October 2002.

[2] Radu Bogdan Rusu, Wim Meeussen, Sachin Chitta,and Michael Beetz. Laser-based perception for doorand handle identification. In International Conferenceon Advanced Robotics (ICAR), Munich, Germany,06/2009 2009.

[3] M. Inaba. Remote-Brained Robotics: Interfacing AIwith Real World Behaviors. In Proceedings of 1993International Symposium on Robotics Research, 1993.

[4] M. Inaba. Developmental processes inremote-brained humanoids. In Preprint of the 8thInternational Symposium on Robotics Research, pages155–160, 1997.

[5] Ikuo Mizuuchi, Ryosuke Tajima, Tomoaki Yoshikai,Daisuke Sato, Koichi Nagashima, Masayuki Inaba,Yasuo Kuniyoshi, and Hirochika Inoue. TheDesign and Control of the Flexible Spine of a FullyTendon-Driven Humanoid “Kenta”. In Proceedings ofthe 2002 IEEE/RSJ International Conference on IntelligentRobots and Systems, pages 2527–2532, 2002.

[6] Ikuo Mizuuchi, Shigenori Yoshida, Masayuki Inaba,and Hirochika Inoue. The development and controlof the flexible-spine of a human-form robot. AdvancedRobotics, 17(2):179–196, 2003.

[7] Ikuo Mizuuchi, Yuto Nakanishi, Yuta Namiki,Tomoaki Yoshikai, Yoshinao Sodeyama, TamakiNishino, Junichi Urata, and Masayuki Inaba.Realization of standing of the musculoskeletalhumanoid kotaro by reinforcing muscles. InProceedings of 2006 6th IEEE-RAS InternationalConference on Humanoid Robots(Humanoids’06), 122006.

[8] Yoshinao Sodeyama, Ikuo Mizuuchi, TomoakiYoshikai, Yuto Nakanishi, and Masayuki Inaba.A shoulder structure of muscle-driven humanoidwith shoulder blades. In Proceedings of the 2005IEEE/RSJ International Conference on Intelligent Robotsand Systems, pages 1077–1082, 2005.

[9] Ikuo Mizuuchi, Hironori Waita, Yuto Nakanishi,Tomoaki Yoshikai, Masayuki Inaba, and HirochikaInoue. Design and Implementation of ReinforceableMuscle Humanoid. In Proceedings of the 2004IEEE/RSJ International Conference on Intelligent Robotsand Systems, pages 828–833, 2004.

[10] Ikuo Mizuuchi, Yuto Nakanishi, Yuta Namiki,Tomoaki Yoshikai, Yoshinao Sodeyama, TamakiNishino, Junichi Urata, and Masayuki Inaba.Realization of standing of the musculoskeletalhumanoid kotaro by reinforcing muscles. In 2006 6thIEEE-RAS International Conference on Humanoid Robots(Humanoids 2006), pages 176–181, Dec 2006.




www.intechopen.com

[11] Ikuo Mizuuchi, Yuto Nakanishi, Yoshinao Sodeyama,Yuta Namiki, Tamaki Nishino, Naoya Muramatsu,Junichi Urata, Kazuo Hongo, Tomoaki Yoshikai, andMasayuki Inaba. An Advanced MusculoskeletalHumanoid Kojiro. In Proceedings of InternationalConference on Humanoid Robots(Humanoids’07), 122007.

[12] Junichi Urata, Toshinori Hirose, Yuta Namiki, YutoNakanishi, Ikuo Mizuuchi, and Masayuki Inaba.Thermal control of electrical motors for high-powerhumanoid robots. In Proceedings of the 2008IEEE/RSJ International Conference on Intelligent Robotsand Systems, pages 2047–2052, 2008.

[13] Yuto Nakanishi, Yuta Namiki, Junichi Urata,Ikuo Mizuuchi, and Masayuki Inaba. Design oftendon driven humanoidAfs lower body equippedwith redundant and high-powered actuators.In Proceedings of the 2007 IEEE/RSJ InternationalConference on Intelligent Robots and Systems, pages3623–3628, Oct 2007.

[14] Yoshinao Sodeyama, Tomoaki Yoshikai, TamakiNishino, Ikuo Mizuuchi, and Masayuki Inaba. Thedesigns and motions of a shoulder structure with awide range of movement using bladebone-collarbonestructures. In Proceedings of the 2007 IEEE/RSJInternational Conference on Intelligent Robots andSystems, pages 3629–3634, Oct 2007.

[15] Rainer Bischoff, Johannes Kurth, Gunter Schreiber,Ralf Koeppe, Alin Albu-Schaffer, Alexander Beyer,Oliver Eiberger, Sami Haddadin, Andreas Stemmer,Gerhard Grunwald, and Gerhard Hirzinger. Thekuka-dlr lightweight robot arm - a new referenceplatform for robotics research and manufacturing.In accepted at International Symposium on Robotics(ISR2010), 2010.

[16] Masahiko Osada Masahiko, Nobuyuki Ito, YutoNakanishi, and Masayuki Inaba. Realizationof flexible motion by musculoskeletal humanoidkojiro with add-on nonlinear spring units. Inaccepted at International Conference on HumanoidRobots(Humanoids’10), 12 2010.

[17] Ikuo Mizuuchi, Takeshi Matsuki, Masayuki Inaba,and Hirochika Inoue. Simulation environment ofhumanoid robot which contains soft structure, andits application to real robot’s motion generation.In Proc. of the 1999 JSME Conference on Roboticsand Mechatronics (ROBOMRC’99), pages 2A1–47–085,1999.

[18] Shigenori Yoshida, Ikuo Mizuuchi, Masayuki Inaba,Yasuo Kuniyoshi, and Hirochika Inoue. Acquisitionof crawl-motion of the flexible-spine humanoid robotby ga. In ROBOMEC’02, pages 2P2–L04, 6 2002.

[19] Ikuo Mizuuchi, Tomoaki Yoshikai, Daisuke Sato,Shigenori Yoshida, Masayuki Inaba, and HirochikaInoue. Behavior developing environment for thelarge-dof muscle-driven humanoid equipped withnumerous sensors. In Proceedings of the 2003 IEEEInternational Conference on Robotics & Automation,pages 1940–1945, 2003.

[20] Kazuo Hongo, Yuto Nakanishi, Yuta Namiki, IkuoMizuuchi, and Masayuki INABA. Reflective steppingwith 6-axis sensors by musculo-skeletal humanoid

legs. In Proceedings of the 25th Annual Conference of theRobotics Society of Japan, page 3H26, 9 2007.

[21] Kazuo Hongo, Yuto Nakanishi, Yuta Namiki,Ikuo Mizuuchi, and Masayuki Inaba. Automaticparameter adjustment of reflexive walking of amusculo-skeletal humanoid. In Proceedings of the2008 IEEE-RAS International Conference on HumanoidRobots, pages 16–21, 12 2008.

[22] Yuto Nakanishi, Ikuo Mizuuchi, Tomoaki Yoshikai,Tetsunari Inamura, and Masayuki Inaba. Pedalingby a redundant musculo-skeletal humanoid robot.In Proc. of International Conference on HumanoidRobots(Humanoids’05), 12 2005.

[23] Kenichi Narioka and Koh Hosoda. Designingsynergistic walking of a whole-body humanoiddriven by pneumatic artificial muscles: An empiricalstudy. Advanced Robotics, 22(10):1107–1123, 2008.

[24] Yuto Nakanishi, Yuta Namiki, Kazuo Hongo,Junichi Urata, Ikuo Mizuuchi, and Masayuki Inaba.Realization of large joint movement while standingby a musculoskeletal humanoid using its spineand legs coordinately. In Proceedings of the 2008IEEE/RSJ International Conference on Intelligent Robotsand Systems, pages 205–210, 9 2008.

[25] Yuto Nakanishi, Tamon Izawa, Tomoko Kurotobi,Junichi Urata, Kei Okada, and Masayuki Inaba.Achievement of complex contact motion withenvironment by musculoskeletal humanoid usinghumanlike shock absorption strategy. In Proceedingsof the 2011 IEEE/RSJ International Conference onIntelligent Robots and Systems, pages 1815–1820, 102012.

[26] Yuto Nakanish, Kazuo Hongo, Ikuo Mizuuchi, andMasayuki Inaba. Joint proprioception acquisitionstrategy based on joints-muscles topological maps formusculoskeletal humanoids. In Proceedings of the 2010IEEE International Conference on Robotics & Automation,pages 1727–1732, 5 2010.

[27] Ryosuke KAGEYAMA, Koichi NAGASHIMA,Hiroaki YAMAGUCHI Satoshi KAGAMI, , MasayukiINABA, and Hirochika INOUE. 3-axis angle sensorusing magnetic coils. In ROBOMECH ’00, pages2A1–54–068, 2000.

[28] Junichi Urata, Yuto Nakanishi, Akihiko Miyadera,Ikuo Mizuuchi, Tomoaki Yoshikai, and MasayukiInaba. A three-dimensional angle sensor for aspherical joint using a micro camera. In Proceedings ofThe 2006 IEEE International Conference on Robotics andAutomation, May 2006.

[29] Ikuo Mizuuchi, Tomoaki Yoshikai, Yuto Nakanishi,Yoshinao Sodeyama, Taichi Yamamoto, AkihikoMiyadera, Tuomas Niemelä, Marika Hayashi,Junichi Urata, and Masayuki Inaba. Developmentof Muscle-Driven Flexible-Spine Humanoids. InProceedings of the 2005 IEEE-RAS InternationalConference on Humanoid Robots, pages 339–344, 2005.

[30] Masayuki Inaba, Ikuo Mizuuchi, Ryosuke Tajima,Tomoaki Yoshikai, Daisuke Sato, Koichi Nagashima,and Hirochika Inoue. Building spined muscle-tendonhumanoid. In Robotics Research: The Tenth InternationalSymposium, pages 113–130. Springer Verlag, 2003.




www.intechopen.com

[31] Kazuo Hongo, Mariko Yoshida, Yuto Nakanishi, IkuoMizuuchi, and Masayuki Inaba. Development ofBilateral Wearable Device ”Kento” for Control Robotsusing Muscle Actuator Modules. In Proceedings of the18th IEEE International Symposium on Robot and HumanInteractive Communication, pages 897–902, 2009.

[32] Fumio Kanehiro, Ikuo Mizuuchi, Youhei Kakiuchi,Masayuki Inaba, and Hirochika Inoue. Developmentof the remote-brained robot which has a nervoussystem by a on-body lan. In Proc. of Annual Conf. ofRobotics Society of Japan, pages 1023–1024, 1997.

[33] Akihiko MIYADERA, Ikuo MIZUUCHI, andMasayuki INABA. The control and communicationsystem for the super large-dof humanoid. InProceedings of the 23th Annual Conference of the RoboticsSociety of Japan, page 1E35, September 2005.

[34] Yuto NAKANISHI, Yuta NAMIKI, Tamaki NISHINO,Junichi URATA, Ikuo MIZUUCHI, and MasayukiINABA. A visualization system for sensor integrationof musculo-skeletal humanoids. In ROBOMEC’08,pages 2A1–D05, 6 2008.

[35] Yuto Nakanishi, Hironori Mizoguchi, TomokoKurotobi, Yuriko Kakehashi, Yoshito Ito, TamonIzawa, Takuma Shirai, Masahiko Osada, TakumiKobayashi, Nobuyuki Ito, Shigeki Ohta, JunichiUrata, and Masayuki Inaba. Body informationdatabase for musculoskeletal humanoids byauto-generation of motion events’ indexes. InProceedings of the 11th SICE System Integration DivisionAnnual Conference, pages 1G2–3, 12 2010.

[36] I. A. Kapandji. Physiology of the Joints. ChurchillLivingstone, 1986.

[37] Appolinaire C. Etoundi, Ravi Vaidyanathan, andStuart C. Burgess. A bio-inspired condylar hinge jointfor mobile robots. pages 4042–4047, 2011.

[38] Kenji Kaneko, Fumio Kanehiro, MitsuharuMorisawa, Kazuhiko Akachi, Go Miyamori, AtsushiHayashi, and Noriyuki Kanehira. HumanoidRobot HRP-4 -Humanoid Robotics Platform withLightweight and Slim Body-. pages 4400–4407, 2011.




www.intechopen.com

design approach of biologically-inspired musculoskeletal … · 2016-10-04 · yuto nakanishi,...

Documents