Limb-Vector Paralleling: A General Approach to Translate Postures from Human to Humanoid Robots

Size: px

Start display at page:

Download "Limb-Vector Paralleling: A General Approach to Translate Postures from Human to Humanoid Robots"

Norman Hicks
5 years ago
Views:

1 Limb-Vecor Parallelig: A Geeral Approach o Traslae Posures from Huma o Humaoid Robos Zhuo-Li Bao, Shih-Li Che, Chi-Chag Tug, Rayi Yau Tara, Hadziq Fabroyir ad Wei-Chug Teg Absrac evelopig iuiive ierface o corol humaoid robos wih high degrees-of-freedom (OF i real-ime remais as a esseial issue. Sice mos humaoid robos have a leas siee OF, achievig real-ime performace i corollig humaoid robo posure is hard o accomplish. This paper proposes a mehod o calculae joi agles of humaoid robos accordig o huma's joi locaios racked by a moio capure sysem. The objecive is o make he humaoid robo imiaig huma posure i real-ime. Our proposed mehod requires a huma moio capure sysem ha could provide a leas hiree joi posiios: a pair of arm (shoulder, elbow, had, a pair of leg (hip, kee, akle, ad a orso. Every joi ha iercoecs wo limbs has o have wo OF. Our proposed mehod is implemeed o RoboBuilder humaoid robo. The eperimeal resuls show ha he robo ca imiae huma posure i real-ime correcly. Moreover, he resuls verify ha he proposed mehod is applicable o differe ype of humaoid robos usig WeBos simulaio sofware. I. INTROUCTION Learig ad mimickig huma s posure is oe of he ieresig research issues i humaoid robos research. Physically desiged accordig o ahropomeric, humaoid robos is mosly desiged wih large umber of OF ad aimed for havig he abiliy i mimickig huma s posure ad moio characerisics. Effors o mimickig huma moio ofe deped o he availabiliy, aalysis, ad qualiy of huma moio daa. The daa is usually obaied usig huma moio capure sysem. Nowadays, rece huma moio capure echologies efficaciously ease researches o mimickig huma s posure io humaoid robos. Geerally, mimickig huma s posure io humaoid robo ivolves several seps: Huma moio iformaio acquisiio, Moio iformaio processig, ad Iverse kiemaic calculaio. The compleiies bewee hese seps are varied ad depede wih he employed mehods ad equipme. As a eample, if he acquisiio process resuls i he form of image, here will be era processig eeded o he moio iformaio-processig sep o rasform he image io joi coordiaes. Therefore, i is impora o oice he employed processig compleiy, due o is effec o compuaio laecy, whe aimig for real-ime performace. Huma moio iformaio ca be acquired hrough differe approaches, wheher as pre-recorded daa [] [8] or Zhuo-Li Bao, Shih-Li Che, Chi-Chag Tug, Rayi Yau Tara, Hadziq Fabroyir, ad Wei-Chug Teg* are wih Neworked Roboics Laboraory, ep. of Compuer Sciece ad Iformaio Egieerig, Naioal Taiwa Uiversiy of Sciece ad Techology, Taipei, Taiwa ( *: Weichug@csie.us.edu.w real-ime daa [9] []. Older moio capure sysem works wih he marker-based sysem ha aached o a huma body [], [9], [0], []. This kid of sysem is cosidered o be icoveie for user ad more epesive. Wih he icreased capabiliy of curre compuig echology, marker-less moio capure sysem [], [0] [] gais more populariy i he rece years. Marker-less moio capure sysem mosly based o he visio sysem, which provides huma moio iformaio i he form of image [], [9], [0], [] ad should be processed laer o obai he moio iformaio. Sophisicaed ye epesive marker-less moio capure sysems could direcly provide huma moio iformaio i fleio agles [], [], [4] hus probably deliver more ease for user o uilize he daa. Several approaches have bee uilized for calculaig iverse kiemaic o rasform he obaied huma moio iformaio io robo moio ad posure. A approach i [] iroduces he use of Maser Moor Map (MMM o rasform fleio agles io robo posiio. Previous researches usually ried o solve iverse kiemaic by usig eural ework [], [6], [9], regressio ad learig [5], ad opimizaio mehod [], [], [8], [0]. iffere mehod was also preseed i [] o solve iverse kiemaic. Obviously, as he umber of robo s OF ges larger, he compleiy of iverse kiemaic calculaio is icreased. Regardless of he various echiques o imiae huma posure io humaoid robo, mosly preseed mehods are oly suiable for specific robo model. A addiioal ask (i.e., ediig moio daa mus be performed if he previously preseed mehods will be implemeed i differe model of humaoid robo. Thus, i is very icoveie for he user ad i will o guaraee similar qualiy resuls. I order o address he limiaio, we propose a Limb-Vecor Parallelig mehod i his paper. Our proposed mehod uilize low-cos ad marker-less huma moio rackig sysem (e.g., Kiec [6] o corol humaoid robo i real-ime, which is applicable for differe ype of humaoid robos. II. LIMB-VECTOR PARALLELING METHO Humaoid robos ca mimic user's moio hrough he limb-vecor parallelig mehod. Each user's limb-vecor correspods o each robo's limb-vecor. The mehod is devoed o check wheher he correspodig limb-vecors are parallel or o. However, he mehod igores he ed effecor ad he posiio of each joi. A. Moio Trackig Sysem ad Robo Requireme I his paper, he moio rackig sysem uses he Microsof Kiec o capure he user s joi iformaio as show i

2 represe he local coordiae ais; ad he X o represe he projecio value of he local -ais o he global X-ais; ad so o. Furhermore, we pu he values io a mari, so we ca beefi from he mari ier produc operaio. P c Figure. P orieaio mari. X-ais (red, Y-ais (yellow, ad Z-ais (gree Table I. I addiio, he humaoid robo should have oe ruk, wo arms, ad wo legs. Furhermore, each joi mus have more ha wo OF (egree of Freedom. TABLE I. Figure. Joi posiio JOINT POSITION AN NAME Joi I Joi ame Joi posiio P Truk [ y z] T P Righ shoulder [ y z] T P Righ arm [ y z] T P4 Righ had [4 y4 z4] T P5 Lef shoulder [5 y5 z5] T P6 Lef arm [6 y6 z6] T P7 Lef had [7 y7 z7] T P8 Righ hip [8 y8 z8] T P9 Righ kee [9 y9 z9] T P0 Righ leg [0 y0 z0] T P Lef hip [ y z] T P Lef kee [ y z] T P Lef leg [ y z] T B. efiiio of Orieaio Mari ad Roaio Mari Acually, orieaio mari ad roaio mari ca be he same hig. Whe he mari is reaed as a orieaio mari, i disiguishes ie projecio values accordig o he local coordiae sysem o describe objec orieaio i he global coordiae sysem. However, whe he mari is cosidered as a roaio mari, i represes a series of roaio operaios. I brief, whe he local coordiae sysem is aliged o he global coordiae sysem, a orieaio mari ca be regarded as a roaio mari, or he oher way aroud. The projecio of he local coordiae sysem a he global coordiae sysem geeraes ie values. We le X, Y, ad Z o represe he global coordiae ais; ad, y, ad z o C. Limb-Vecor Calculaio Orieaio Mari Firsly, we calculae P s OM (Orieaio Mari as illusraed i Figure ad Eq. (. The firs colum is he ui vecor P 5 o P. The secod colum is he ui vecor from P o he middle bewee P ad P 5. The hird colum is he ui vecor from firs colum crossig he secod colum. OM Every eleme i orieaio mari is calculaed usig he followig equaio. A imagiary poi P c mus be calculaed firs before calculaig he secod colum of orieaio mari. The hird colum of orieaio mari will be calculaed afer every eleme i he firs ad secod colums of orieaio mari have bee obaied. P c P P 5 P P5 P P 5 5 P5 Pc P P P P c

A se of roaio marices R,,Ri ha correspods o he adjace joi of robo limb-vecor is also eeded sice he ed oupu is he roaio agle of each OF joi. The geeral form of calculaig each joi roaio is showed i Eq.

3 l l Joi Roaio Agle Calculaig each joi roaio agle is he fial sep ha required o parallelizig user ad robo s limb-vecors. I geeral, he calculaio requires curre limb-vecor of he robo V - ad epeced limb vecor of he robo V, which is he same wih correspodig huma limb-vecor. A se of roaio marices R,,Ri ha correspods o he adjace joi of robo limb-vecor is also eeded sice he ed oupu is he roaio agle of each OF joi. The geeral form of calculaig each joi roaio is showed i Eq. 7. Usig his equaio, every joi roaio agle ca be obaied from each roaio mari. R R R V i V Secodly, we calculae each limb-vecor V (X,Y,Z as show i Figure usig Eq. 6. The amig coveio for each vecor is show i Table. V TABLE II. P OM P Vecor ame V V V4 P P VECTOR NAMING CONVENTION Vecor oaio [, y, z,] T [, y, z,] T [4, y4, z4,] T T For eample, robo s righ-arm has wo OF jois o he shoulder ad wo OF joi i he elbow. Figure shows he illusraio for robo s righ-arm, is correspodig limb-vecor ad joi. The firs wo OF jois, which correspod o upper arm, are joi roaio o z-ais γ ad joi roaio o -ais α. The secod wo OF joi, which correspod o lower arm, are joi roaio o y-ais ß ad joi roaio o -ais α. Thus, o make robo ad huma s limb-vecor become parallel, he geeral form i Eq. 7 is he used for calculaig each joi roaio agle o robo s righ-arm, as show i Eq.8. R R Rz V V R R R V V z y V5 V6 V7 V8 V9 V0 V V V [5, y5, z5,] T [6, y6, z6,] T [7, y7, z7,] T [8, y8, z8,] T [9, y9, z9,] T [0, y0, z0, ] T [, y, z,] T [, y, z,] T [, y, z,] T Usig Eq.9, every correspodig joi agle ca be calculaed by dissec every correspodig mari elemes i Eq.8 io ilie equaio. As show i Fig., robo s righ-arm limb-vecor V - ad V - are [, 0, 0] T, ad he epeced limb-vecor V is [, y, z,] T ad V is [, y, z,] T. Thus, afer he dissecio process, he followig equaio ca be obaied ad used o calculae each joi roaio agle for righ-arm. (a (b (c (d Figure. Illusraio for righ arm limb-vecor calculaio. Upper arm limb-vecor ad is joi (a, upper arm correspodig limb-vecor (b, lower arm limb-vecor ad is joi (c, ad lower arm correspodig limb-vecor (d.

4 arccos( arca( z arccos( arca( y,,,, y, z I geeral, he calculaio process is similar for calculaig he oher robo s limb-vecor (i.e., lef arm, righ leg, ad lef leg. All joi roaio agles are always calculaed usig he geeral form ha show i Eq. 7. III. EXPERIMENT A. Purposes ad Resuls Verificaio To verify he correcess, we apply wo evaluaios. Firs, we observe he cosisecy of moio bewee user ad robo. Secod, we deermie robo's posures righess by calculaig he jois' agle. Furhermore, he laecy moio mimic bewee robo ad user is measured. The mehod's algorihm is cosidered as efficie whe he robo is mimickig he user's moio wihi a small delay. All humaoid-robos provided by he simulaor are able o mimic user moio cosisely. However, he mimickig process is coversely o cosise due o he followig codiios: The daa from Kiec are probably usable, ad he deecio of ais-posiio s variable resisor value may coai errors. Thus, he robos servo roaio agles may o be accurae. The robos cao ge reliably fas compuers, adequae serial commuicaio baud raes, real-ime operaig sysems, or rapid algorihms o suppor heir special eeds regardig reacio rae. B. Sysem Archiecure Sice our eperime is coduced usig various robos, we adop Azad e.al. framework [] i our sysem. Their framework provides he capabiliies o perceive, visualize, reproduce, ad recogize huma moio. I heir framework s core, hey also defied a referece kiemaic model he Maser Moor Map. They showed ha heir approach performs well by havig a buil-i marker-less huma moio capure sysem. Our sysem adds a coversio from huma moio capure from Kiec o BVH files. Aferwards, hese files are uilized o corol various robos. IV. ISCUSSION A mehod ha uses geeral approach i raslaig posure from huma o humaoid robo posure is preseed i his paper. The preseed mehod is called Limb Vecor Parallelig. Limb Vecor Parallelig provides easy o use ad accurae algorihm for corollig humaoid robo posure usig huma moio daa. The eperime video of how his mehod work ca be viewed i his followig lik: hp://rlab.csie.us.edu.w/lvp., Compared o oher mehods ha mosly discuss he relaio bewee joi agle ad proimiy of ed effecor posiio (e.g., iverse kiemaic, his mehod is more focus o aalyzig he vecor of huma limbs. Our proposed mehod provides mimickig accuracy of 00% as log as he give se of rules is fulfilled. Moreover, sice he compuaio compleiy is liear, his mehod ca be used i real ime implemeaio. From he implemeaio of Limb-vecor Parallelig o humaoid robo, we fid ou several impora characerisics ha show he advaage ad disadvaage of our mehod. The characerisics are: Our mehod successfully imiae huma posure alhough he huma posiio is o eacly i perpedicular wih he moio capure sysem. The posiio of ed effecor is ofe icorrec. The laecy i humaoid robo made he overall performace seems uresposive, which is probably caused by he robo corol processig ui. There is a jier i he moio daa, which is probably caused by he low-cos moio capure sysem. The humaoid robo is ofe fall whe performig some gesure. ue o he above-meioed codiio, we are plaig o add balace corol ad fall deecio sysem i he fuure. Implemeig our mehod wih more sophisicaed moio capure sysem is also cosidered o observe he relaio bewee overall sysem performace ad moio daa qualiy. REFERENCES [] A. Billard ad M. J. Maarić, Learig huma arm movemes by imiaio:: Evaluaio of a biologically ispired coeciois archiecure, Roboics ad Auoomous Sysems, vol. 7, o., pp , 00. [] A. Ude, C. G. Akeso, ad M. Riley, Programmig full-body movemes for humaoid robos by observaio, Roboics ad Auoomous Sysems, vol. 47, o., pp. 9 08, 004. [] M. Ruchaurucks, S. Nakaoka, S. Kudoh, ad K. Ikeuchi, Humaoid robo moio geeraio wih sequeial physical cosrais, i Roboics ad Auomaio, 006. ICRA 006. Proceedigs 006 IEEE Ieraioal Coferece o, 006, pp [4] A. Nakazawa, S. Nakaoka, K. Ikeuchi, ad K. Yokoi, Imiaig huma dace moios hrough moio srucure aalysis, i Iellige Robos ad Sysems, 00. IEEE/RSJ Ieraioal Coferece o, 00, vol., pp vol.. [5] N. S. Pollard, J. K. Hodgis, M. J. Riley, ad C. G. Akeso, Adapig huma moio for he corol of a humaoid robo, i Roboics ad Auomaio, 00. Proceedigs. ICRA 0. IEEE Ieraioal Coferece o, 00, vol., pp vol.. [6] A. Safoova, N. Pollard, ad J. K. Hodgis, Opimizig huma moio for he corol of a humaoid robo, Proc. Applied Mahemaics ad Applicaios of Mahemaics, vol. 78, 00. [7] S. Nakaoka, A. Nakazawa, K. Yokoi, H. Hirukawa, ad K. Ikeuchi, Geeraig whole body moios for a biped humaoid robo from capured huma daces, i Roboics ad Auomaio, 00. Proceedigs. ICRA 0. IEEE Ieraioal Coferece o, 00, vol., pp vol.. [8] W. Suleima, E. Yoshida, F. Kaehiro, J.-P. Laumod, ad A. Moi, O huma moio imiaio by humaoid robo, i Roboics ad Auomaio, 008. ICRA 008. IEEE Ieraioal Coferece o, 008, pp

5 [9]. Masui, T. Miao, K. F. Macorma, ad H. Ishiguro, Geeraig aural moio i a adroid by mappig huma moio, i Iellige Robos ad Sysems, 005. (IROS IEEE/RSJ Ieraioal Coferece o, 005, pp [0] M. o, P. Azad, T. Asfour, ad R. illma, Imiaio of huma moio o a humaoid robo usig o-liear opimizaio, i Humaoid Robos, 008. Humaoids h IEEE-RAS Ieraioal Coferece o, 008, pp [] Y.-A. Zheg, Humaoid Robo Behavior Emulaio ad Represeaio, 0. [] F. Zuher ad R. Romero, Recogiio of Huma Moios for Imiaio ad Corol of a Humaoid Robo, i Roboics Symposium ad Lai America Roboics Symposium (SBR-LARS, 0 Brazilia, 0, pp [] P. Azad, T. Asfour, ad R. illma, Toward a Uified Represeaio for Imiaio of Huma Moio o Humaoids, i Roboics ad Auomaio, 007 IEEE Ieraioal Coferece o, 007, pp [4] T. Liu, H. Usuomiya, Y. Ioue, ad K. Shibaa, Sychroous imiaio corol for biped robo based o wearable huma moio aalysis sysem, i Iellige Robos ad Sysems, 008. IROS 008. IEEE/RSJ Ieraioal Coferece o, 008, pp [5] A. Souza, S. Vijayakumar, ad S. Schaal, Learig iverse kiemaics, i Iellige Robos ad Sysems, 00. Proceedigs. 00 IEEE/RSJ Ieraioal Coferece o, 00, vol., pp vol.. [6] Microsof Corporaio, Kiec for Widows. [Olie]. Available: hp:// [Accessed: 07-Ja-04].

Introduction to Orthogonal Frequency Division Multiplexing (OFDM)

Introduction to Orthogonal Frequency Division Multiplexing (OFDM) Wireless Iformaio Trasmissio Sysem Lab. Iroducio o Orhogoal Frequecy Divisio Muliplexig OFDM Isiue of Commuicaios Egieerig aioal Su Ya-se Uiversiy OFDM Overview OFDM Sysem Model Orhogoaliy Oulie Muli-carrier