ICAMechS The Navigation Mobile Robot Systems Using Bayesian Approach through the Virtual Projection Method

Transcription

1 ICAMechS 2012 Advanced Inelligen Conrol in Roboics and Mecharonics The Navigaion Mobile Robo Sysems Using Bayesian Approach hrough he Virual Projecion Mehod Tokyo, Japan, Sepember 2012 Luige VLADAREANU, Gabriela TONT, Vicor VLADAREANU, Florenin SMARANDACHE and Lucian CAPITANU 1

2 Improvemen of Dynamical Sabiliy for he Real Time Walking Robo Conrol PERO 1. Inroducion 2. Dynamic Sabiliy Conrol 3. Simulaneous Localizaion and Maping 4. A Peri Ne and Markov Chains Approach 5. Robo simulaion hrough Peri Nes 6. Virual projecion archiecure for robo experimenal conrol 7. Resuls and Conclusions 2

3 1. Inroducion The Navigaion and Sabiliy of he Mobile Robo Sysems The approach of he localizaion and navigaion problems of a mobile robo which uses a WSN which comprises of a large number of disribued nodes wih cameras as main sensor Hybrid force posiion conrol hrough operaional space mehod The Zero Movemen Poin (ZMP) mehod by processing inerial informaion of force orque and iling and by implemening inelligen high level algorihms 3

4 1. Inroducion The Navigaion and Sabiliy of he Mobile Robo Sysems Walking robo conrol sraegy is based on hree approaches: real ime balance conrol walking paern conrol predicable moion conrol in correlaion wih a sochasic model of assessing sysem probabiliy of unidirecional or bidirecional ransiion saes 4

5 1. Inroducion The Navigaion and Sabiliy of he Mobile Robo Sysems he sysem archiecure conrol was compleed for HFPC walking robos (hybrid force posiion conrol) wih wo oher conrol funcions Online conrollers were designed for he objecives of he conrol schemes and planned in he early sages of walking. Was obained he walking conrol algorihm wih online simulaions obained experimenally hrough he virual projecion mehod. 5

6 2. Dynamic Sabiliy Conrol 6 Dynamic conrol of he walking robos moion

7 2. Dynamic Sabiliy Conrol Real ime balance conrol of he robo using sensorial feedback, has 3 online ypes of conrol loops: Damping conrol sraegy aims o eliminae he oscillaions ha occur in he single suppor phase. ZMP compensaion conrol sraegy. Conrol sraegy consiss in mahemaical modeling of ZMP compensaor hrough he springloaded invered pendulums. Gai iming conrol sraegy. Generaes a signal o block he walking diagram if he foo doesn ouch he ground a he end of phase 2 and 4. 7

8 2. Dynamic Sabiliy Conrol The walking paern conrol, may be changed periodically according o he informaion received from an ineria ransducer during each walking cycle. Modeling of plaform balance. In erms of conrol sraegy, periodically adjus he laeral balance ampliude of he plaform in order o move ZMP, correced by measuring ZMP during each walking cycle. Roaion/advance plaform conrol sraegy, allows he cenral posiion of he plaform o move in he opposie direcion o he inclined ransverse plane so ha he swinging movemen can be well balanced. 8

9 2. Dynamic Sabiliy Conrol Predicable moion conrol. Is based on generaing probable robo movemens achieved by processing earlier movemens, in order o avoid overhrowing he walking robo. Conrol of landing posiion, is for o compensae foo landing posiion on ground, in order o walk owards he direcion of he fall, by a wising movemen of he plaform. Conrol of il over he side of safey of he robo. The il conrol loop prevens he fall of he walking robo in laeral direcions, in case of moving on a bumpy field or exernal forces. 9

10 3. Simulaneous Localizaion and Maping Auonomous mobile robos sysems ha can perceive heir environmens, reac o unforeseen circumsances, and plan dynamically in order o achieve heir mission have he objecive of he moion planning and conrol problem. The relaionship beween he subasks mapping and modeling of he environmen; pah planning and selecion; pah raversal and collision avoidance ino which he navigaion problem is decomposed, is shown in diagram. 10

11 3. Simulaneous Localizaion and Maping Moion planning of mobile walking robos in uncerain dynamic environmens based on he behavior dynamics of collision-avoidance is ransformed ino an opimizaion problem. Applying consrains based on conrol of he behavior dynamics, he decision-making space of his opimizaion. In his sense, a new conrol algorihm has been sudied and analyzed for dynamic walking of robos based on sensory ools such as force / orque and inerial sensors. Disribued conrol sysem archiecure was inegraed ino he HFPC archiecure so ha i can be conrolled wih high efficiency and high performance. 11

12 4. A Peri Ne and Markov Chains Approach Peri Nes The ineres in modeling by means of PN came from he following reasons: More deailed causal modeling considering he wo main aspecs; informaion flow wihin he sysem akes he form of process and ask and conrol sequences, may saisfy he Markovian propery [10, 17]; allow o build dynamic models which incorporae ime informaion of he process developmen ; sequencing and planning acions can be checked and moniored hroughou sysem saes ha can be relaed o insecuriy condiions.

13 4. A Peri Ne and Markov Chains Approach Peri Nes Peri nes share many common properies wih oher formalisms. Peri nes can normally be convered o ime Peri nes for simulaion and performance modeling. Peri nes have a dual ideniy. They can be represened graphically and non graphically; Compared wih oher formalisms Peri nes are preferable for visualizaion and comprehension by differen sakeholders.

14 4. A Peri Ne and Markov Chains Approach Peri Nes Peri nes have been used o model: i) hardware; ii) sofware sysems; iii) communicaion sysems; iv) manufacuring v) sofware modeling.

15 4. A Peri Ne and Markov Chains Approach Peri Nes In order o perform he evaluaion Peri ne, i is necessary o quanifying he probabiliies of saes using Markov chains, assuming following hypohesis: i) sabiliy and insabiliy of he componen is random, he currenly sae depends only on he immediaely preceding sae; ii) he insabiliy inensiy, and sabiliy inensiy are ime consan.

16 4. A Peri Ne and Markov Chains Approach Peri Nes Assimilaing developmen over ime of sysem hrough he various saes ha may occur as a resul of fail and resore of elemens wih a coninuous ime Markov process, solving is done by he sysem of differenial equaions, wrien under generalized form as he marix

17 4. A Peri Ne and Markov Chains Approach Peri Nes Modeling he saes wih possible ransiions for robo

18 4. A Peri Ne and Markov Chains Approach Peri Nes Sysem Peri Ne Represenaion

19 4. A Peri Ne and Markov Chains Approach Peri Nes Associaed Markov graph of he sysem modeled by means of Peri nes

20 4. A Peri Ne and Markov Chains Approach Peri Nes Sae probabiliy vecor P () sysem and marix of ransiion raes C ()

21 Peri Nes 4. A Peri Ne and Markov Chains Approach The sae equaions of he sysem () () () () [ ] () () () () () () () () () () [ ] () () () () () () () () [ ] () () () () () () () () () () [ ] + + = + + = + + = = P P P P P P P P P P P P P P P P / / / / 1 μ μ λ λ μ μ λ λ μ μ λ λ μ μ λ λ

22 4. A Peri Ne and Markov Chains Approach Peri Nes The funcionaliy funcions Fu () and insananeous availabiliy A () vecors

23 5. Virual Projecion Archiecure In order o verify he force posiion dynamic conrol performance of walking robos, in addiion o inegraing he dynamic conrol loops hrough he ZMP mehod, he virual projecion sysem archiecure allows new conrol funcions simulaneous localizaion and mapping (SLAM) inerface srucural vibraion conrol funcions of he operaing mechanism inerface cooperaion funcions for obsacles avoidance inerfece 23

24 5. Virual Projecion Archiecure The virual projecion mehod realizes he dynamic force posiion conrol performance esing by inegraing he fuzzy muli sage mehod wih resolved acceleraion in conrolling he dynamic force posiion conrol loops hrough ZMP mehod and anoher wo conrol loops. Mehod is paened by he auhors 24

25 5. Virual Projecion Archiecure 25 A number of n funcions of he conrol inerfaces ICF1-ICFn ensure he developmen of open archiecure conrol sysem by inegraing a number of n conrol funcions Addiional funcions provided by he mecharonic conrol : rajecory racking, walking conrol scheme for ripod robos, cener of graviy conrol, and orienaion conrol by image processing

26 5. Virual Projecion Archiecure The HFPC sysem was designed wih an open archiecure (OAH) in a disribued and decenralized srucure o enable easy developmen of new applicaions or adding new hardware modules or sofware for new conrol funcions. As basic funcions here are: Hybrid force-posiion conrol hrough operaional space mehod Complian conrol hrough muli-sage fuzzy mehod Damping conrol sraegy ZMP compensaion conrol sraegy Simulaneous Localizaion and Mapping sraegy 26

27 5. Virual Projecion Archiecure Hybrid force-posiion conrol using operaional space mehod, allow dividing a vecor wih he ransformaion marix A in wo perpendicular componens hrough a se of associaed projecion marices. Complian conrol hrough muli-sage fuzzy mehod allows, by using resolved acceleraion conrol, geing a conrol sysem in which dynamic and kinemaic sabiliy are simulaneously achieved in rigid environmens. 27

28 5. Virual Projecion Archiecure ModBus RS485, RS232 Ms PLC TC G.R.M.R. 2* θi (i=3) 6* θi (i=3) S.C.C. θ i - > Xc δ Xc - > δθi δ f - > δxc C.V. ACSM 2* θi 2*3 SP MA 2* θiref IGR LAN V.P.C. PC Simulaion hrough virual projecion 28

29 6. Robo simulaion hrough Peri Nes Peri Nes

30 6. Robo simulaion hrough Peri Nes Peri Nes

31 7. Resuls and Conclusion View from he righ side camera movemen Peri nes provide a promising soluion owards he developmen quaniaive approach of dynamic discree / sochasic even sysems of ask planning for mobile robos. For a deeper insigh ino conrol and communicaion governing ask assignmen of he robo, he enire discree-even dynamic evoluion of ask sequenial process have o be linguisically described in erms of represenaions. A comprehensive inference framework is required in analyzing he daabase of rules of exper sysems, Peri nes using a large amoun of deails for building analysis, even for small sysems, which lead o high coss. 31

32 7. Resuls and Conclusion Resuls and conclusions. The research evidences ha sable gais can be achieved by employing simple conrol approaches which ake advanage of he dynamics of complian legs. The complian conrol sysem archiecure was compleed wih racking funcions for HFPC walking robos hrough he implemenaion of many conrol loops in differen phase of he walking robo, led o adap he robo walking on sloping land, wih obsacles and bumps. 32

33 7. Resuls and Conclusion Resuls and conclusions. The conrol sysem is disribuive wih muli-processor devices for join conrol, daa recepion from ransducers mouned on he robo, peripheral devices conneced hrough a wireless LAN for off-line communicaions and CAN fas communicaion nework for real ime conrol. Disribued conrol sysem archiecure was inegraed ino he HFPC archiecure so ha i can be conrolled wih high efficiency and high performance. 33

34 7. Resuls and Conclusion View from he righ side camera movemen The resuls obained hrough simulaion and experimens show an increase in mobiliy, sabiliy in real condiions and obaining of high performances relaed o he possibiliy of moving walking robos on errains wih a configuraion as close as possible o real siuaions, respecively developing new echnological capabiliies of he walking modular robo conrol sysems for slope movemen and walking by overaking or going around obsacles. 34

35 ICAMechS 2012 TOKYO, JAPAN, 2012 Acknowledgemens. The auhors wish o express heir graiude o he Research and Educaional Minisry (MEC) and o he Romanian Academy for heir suppor of he program of work repored herein. The work ook place as par of research projec no. 263/ in he framework of Grans of CNCSIS Program IDEAS, PN II (Naional Program for Scienific Research and Innovaion Technologies). HFPC Walking Robos

36 ICAMechS 2012 Advanced Inelligen Conrol in Roboics and Mecharonics Tokyo, Japan, Sepember 2012 Thank you for your aenion Luige VLADAREANU, Gabriela TONT, Vicor VLADAREANU, Florenin SMARANDACHE and Lucian CAPITANU 36