<TITLE: Impedance Control of a Water Hydraulic Manipulator for Teleoperation Applications
ACADEMIC DOMAIN: technology
DISCIPLINE: automation engineering
EVENT TYPE: doctoral defence discussion
FILE ID: UDEFD100
NOTES: continuation of UDEFP100

RECORDING DURATION: 123 min 24 sec

RECORDING DATE: 23.6.2004

NUMBER OF PARTICIPANTS: unknown

NUMBER OF SPEAKERS: 3

S1: NATIVE-SPEAKER STATUS: Spanish, Catalan; ACADEMIC ROLE: research student; GENDER: male; AGE: 24-30

S2: NATIVE-SPEAKER STATUS: Danish; ACADEMIC ROLE: senior staff; GENDER: male; AGE: 51-over

S3: NATIVE-SPEAKER STATUS: Finnish; ACADEMIC ROLE: senior staff; GENDER: male; AGE: 51-over

SU: unidentified speaker

SS: several simultaneous speakers>



<PRESENTATION UDEFP100 by S1>

<S2> first of all i will er say i'm very honoured to be asked to be opponent together with my colleague on this er doctoral dissertation work it's a very very (innovative) for model control aspect combined with operator to remote control other systems , i'm i also want to say that there's a very good from my point of view new research , contribution in this area for using water hydraulic for remote control and in particular using impedance control scheme where it means to regulate dynamics between force and motion of a manipulator , i think the candidate also in this er presentation demonstrated with example where force and motion are related together to control a manipulator with operator , if we look at the doctoral thesis we have a list of thir- er 83 very important reference in research covering research contribution from most of the world including united states japan europe you mention it there's so many key reference in this work and it's a very important area er looking for hazardous application for nuclear power application for a lot of other applications where water hydraulics is the future can er give new contributions for new applications , the candidate have demonstrated good scientific doctoral thesis work and i'm very impressed with the good written thesis in english from my point of view and a conclusion and an outlook for the future highlighted the importance of water hydraulics , combined position and force control and also have er compare with another experimental research carried out at the university here in tampere who is a leading er university from my point view so it's a good opportunity to tell people here in the lecture room that we have very significant research contributed in this huge area focusing on impedance control schemes for manipulators , the (xx) and the doctoral <SIC> thesises </SIC> it looks for also cost aspect so you can using pressure for instance versus force sensors and also this software developed for the communication between the operator and on the other side the manipulator is a very good contribution to this doctoral dissertation work , furthermore i appreciate that the candidate also demonstrated an outlook for the future to look for sixth degree of remote telemanipulation , (over what) manipulators using oil hydraulic and illustrated that the good candidate for this er future telemanipulated could be water hydraulic using rotary vane actuators with servo valve for controlling the teleoperators , it also has some good er contributions for research in the future looking for er intelligent control and mention adaptive set of er impedance parameters for contact and also for non-contact situations in the future , in order to approve maybe minimise (both position) and torque errors so with these comments i will er thank for good contribution in engineering research on water hydraulics for manipulators in the future </S2>
<P:05>
<S3> okay let's go to after this introduction to this thesis , first as a beginning i would like to ask about the title of the work <S1>  mhm-hm </S1> how well it described the work the content of the work could it be a little bit different </S3>
<S1> well i think i wanted to put in the title the main contributions of this thesis which is impedance control applied to water hydraulics and teleoperation also applied to to water hydraulics there have been some er or quite a lot of research with teleoperation and oil and electric manipulators and also impedance control with oil and er electric manipulators but er there wasn't anything er with a water hydraulic manipulator so i wanted to address that er to fill that gap and i think that the title described the , very well the contents of the the book </S1>
<S3> okay yes er i agree there are three three key aspects impedance control water hydraulic manipulator and teleoperation so i think this is very very good title for for the work you <S1> [mhm] </S1> [have] you have done , then some words about the thickness of the book <S1> [mhm] </S1> [it's almost] 150 pages so <S1> [yes] </S1> [a little bit] you was you were not able to say it shorter because to say short it it means that you have to think about more than explaining you can explain things around <S1> [mhm] </S1> [many] with many words but er to say it short and clear that's that it's </S3>
<S1> yeah i i wanted to make it perhaps a bit longer but er clearer than er shorter and a bit confused so that's why i it became a bit longer i wanted to explain things very clearly that's why it became almost 150 pages </S1>
<S3> okay yes er well to be honest it's quite clear and it's quite easy also to follow all what you have you have done you have (xx) the equations quite thoroughly and explained that's that's true and it's it's that's the one of the reasons it's so so big , but anyway to go through 150 pages it will take a lot of time lot of questions so i think next we open open the book , or before opening the book you also explained about this impedance control i think this impedance control is not very common er , according to the name for example i mean pressure control you understand what you are controlling you are controlling the pressure or position control you are controlling the position how do you explain what do you mean by impedance control </S3>
<S1> well </S1>
<S3> you explained that er [when you showed slides] </S3>
<S1> [yeah (to show)] yeah </S1>
<S3> but now now without slides </S3>
<S1> well just to , impose to a control to impose to a mechanical system a desired dynamic behaviour and usually the desired dynamic behaviour is of a critically dumb system and er what we want to do with impedance control is to emulate the behaviour of a desired system it can be second-order system or it can be even first-order but the idea is that er we transform our system into a more simple one which we can manage easier and er the idea is to control or to manage the dynamics of the system </S1>
<S3> okay , then if we open book and say some words about the , nomenclature that you have used there are let's say to me seems to be quite strange nomenclature for example you have this D as a specific volume usually i think V is used as a volume why D , on page eight </S3>
<S1> <SIGH> well i'm er using V for er fixed volume er is er i'm using it for average contained contained volume of each actuator V zero and er this is fixed volume which is not varying and er specific volume is er , is not er a fixed amount of volume it's er , it's kind of a a gain that er gives you , gives you the volume depending on the position of the of the vane , i didn't choose V and then er the choice of D was just er because i had seen it in some oil articles i cannot recall now where but er i have seen it somewhere else </S1>
<S3> okay i thought V is i think it's as good as another letter too but er usually V is used as a specific volume , then about this er kinematics you have used F-K for forward kinematics and I-K for inverse kinematics <S1> mhm </S1> usually T block letter T is used as as de- for describing that transformation matrix </S3>
<S1> er yes , but er , but er , F-K is is a function is more at a conceptual level than er at a pure mathematical level and er , and er you cannot er directly describe inverse kinematics with T you have to solve that matrix in order to get the inverse kinematics and er also i'm i'm using T for some other purposes like torque transfer function or torque and i didn't want to mix mix terms and er since er it's appearing a couple of times only it's not er very used and it's also used in the text it's er i use F-K because it er brings to your mind very quickly in my opinion forward kinematics </S1>
<S3> yeah what is forward kinematics what it does it describe </S3>
<S1> well if you have a manipulator with er several joints er forward kinematics consist- consists of er knowing er what's the position and the orientation of the of the last link of the manipulator if you know the values of the joints that are before that joint </S1>
<S3> and what is the inverse </S3>
<S1> and inverse kinematics means that if you want your manipulator to go to that position and that orientation er which joint values do you need to adjust in the manipulator to achieve such an end location </S1>
<S3> okay yes the or usually forward kinematics is also called direct kinematics and it , a description from joint space to cartesian <S1> [yeah] </S1> [space] and the other one is vice versa , er when , well there are time constant you have used two time constants is there any difference between these on page ten tau time constant and tau with subscript P does it </S3>
<S1> they are just er used in two different contexts i didn't want to put the same er to use the same er letter because it could er mixed up things </S1>
<S3> okay then to the introduction once again i would like to hear your what is considered as a remote control or is it dependent on the distance and er how much the distance has to be if you consider is it as a remote <S1> mhm </S1> how many metres for example is a remote control system </S3>
<S1> well , in theory remote can be very small distance just er some distance but er it depends on on the application if er you are talking about space applications the distance is quite long but er in general there should be some kind of a transmission mechanical or electrical between the er manipulator and the control station and the manipulator should have some kind of a control system which is communicating with the control system at the er operator site , and er usually you want to separate the operator from the from the task because er there is er some problem or there is er er some danger or there is some hazardous er activity happening in the in the remote site </S1>
<P:12>
<S2> well i have to er , we're coming to it about the modelling but you also mention in the nomenclature you have on page nine you have the K with the index Q you call it servo valve coefficient er you think it's the same as very often is called by also (xx) flow [gain] <S1> [flow] gain </S1> and that's the same and what's your argument for using Q for angular position also on page nine very often Q is for fluid (power) @being used@ for flow but </S2>
<S1> yeah i i use a capital Q for [for flow] </S1>
<S2> [yes i] recognised that sorry </S2>
<S1> but er , i could have i was thinking of using theta er greek letter or Q for er angular position but er i i thought that Q was also good and er it's also used in er in robotics perhaps also in nomenclature </S1>
<S2> yes and and then the last er also when you come to the equations you use on page ten er you define omega as the natural frequency er sometimes it's often called in (xx) (control) (xx) at least for putting an index er N for underlining that's the natural frequency </S2>
<S1> oh [it's because] </S1>
<S2> [you use] omega with the index capital N <S1> [okay yeah yeah] </S1> [for the for the] but er also you have defined it clearly so <S1> [mhm] </S1> [(i was just)] thinking about </S2>
<S1> mhm-hm also sometimes <S2> [yeah] </S2> [it's also] with sub-index H to indicate hydraulic natural frequency but er i just chose to to use er just omega </S1>
<S2> yes er i commented one question in the introduction i see that you (change) the introduction er to some of the application areas you mentioned undersea mining and also in space er have you in your research looked for sub-sea applications for offshore being as possible with er robot manipulators <S1> mhm </S1> to assemble and disassemble </S2>
<S1> well we <S2> [(xx)] </S2> [don't have] any we don't have any <S2> [(xx)] </S2> [applications] in in mind but er i think that water hydraulics is a quite er good er technology for underwater because er any external leaks that might happen will not er contaminate the the sea and er in the future er i think it's it's a possibility to have er underwater robots er gov- er some research about underwater robots could be carried out in this institute </S1>
<S2> <WHISPERING> (xx) </WHISPERING> </S2>
<S3> yeah <P:05> er you use or your motivation is this either research concept and er do some remote maintenance tasks there er is this er well how do how do you got or where do you got this this research idea was it er in your interest or was it introduced by by someone else or just that I-T-E-R project happened to be here and </S3>
<S1> mhm well the I-T-E-R project er has been a- at this institute since 94 and er when i came to work i was working i came to work into that project and er i got interested and i got to know water hydraulics and er then one project came er related very closely related with this dissertation so i was already in the project then i just decided that er this would be an interesting er research area and i just joined the the project </S1>
<S3> mhm er also in in the motivation you say that er well you're motivating the use of water hydraulic system that er you also state that er there is a need this for <READING ALOUD> minimum pay payload of fifty kilograms <S1> mhm-hm </S1> maximum reach of 1.5 metres the accuracy of one millimetre at the end effector </READING ALOUD> and you say that the in some place you say that the <READING ALOUD> a manipulator electrically controlled manipulator are not able to handle these kind of of loads </READING ALOUD> er yes <READING ALOUD> the loads to be handled exceed the capacity of manual teleoperation arms </READING ALOUD> but er for example there are there exists such er robots for example kuka has a big robot which is able to handle even fifty kilograms and the reach might be three metres and we have a kuka robot er which is able to handle one 110 kilograms the reach is 2.8 me- met- 2.8 metres and (replicability) under one one millimetre why just to select <S1> [mhm] </S1> [a a] robot and make changes </S3>
<S1> yeah i have to say that i'm not familiar with those robots but er what about the physical size of the robot is it er [large] </S1>
<S3> [quite] quite large [(xx)] </S3>
<S1> [because] er these er these operations take place in (xx) constrained environment where there isn't er much space at all and i think the good properties of hydraulics of compact size actuators are very useful or they are almost the only possibility <S3> [(xx)] </S3> [in that] application </S1>
<S3> that's true these er hyd- er electric mani- industrial manipulators let's call them industrial manipulators they are quite massive <S1> mhm </S1> if they can carry these kind of of loads over one <S1> [mhm yeah] </S1> [hundred kilograms] </S3>
<S1> another another issue is that er magnetic fields in the electric motors might be interfered by radiation that is er in the nuclear reactor so that's another reason <S3> [yeah] </S3> [for] choosing water </S1>
<S3> then er about this water hydraulics systems and hyd- let's say hydraulics systems is usually im- oil is used as a medium in if you think about er the they're not (xx) oil hydraulics that's just like you are saying water hydraulics , but er is it just possible to change the oil with water just put the take the oil oil away from the system and put water </S3>
<S1> well </S1>
<S3> what happens </S3>
<S1> @okay@ er the first thing i have to address is the er compatibility of materials you should have er stainless steel materials if er the oil device is already stainless steel that's one er point for er for being able to use the same device , another device is or another issue is er the one that water viscosity er the viscosity of water er causes more leakage for the same gap so or you have to er change the the sealings or you have to manufacture er mhm parts that are in contact with water with some tighter tolerances to fit er to (radio ) (xx) gaps between between parts , but er that will be the the main issues while replacing oil with with water </S1>
<P:06>
<S3> then er , about these er sensors you you are using pressure sensor as to measure the the forms , er and you also explain why pressure sensors are are used , is it more simple to use to sense the force at the end of the manipulator by using a force sensor or </S3>
<S1> yeah it is more simple but it is er more expensive also it's er it's a trade-off between er cost and performance if you if you can have a bit worse performance with half of the cost then it might be good idea to use pressure transducers and the (radius) performance with pressure transducer is mainly because you have to estimate er friction and gravity forces and compensate them with force sensors you don't have to do that's the main the main advantage </S1>
<S3> yes and and usually if we think about manipulators they they don't like to pay anything extra <S1> [mhm] </S1> [you] you buy er force sensor and but at the end of the arm so that's in many cases too ex- too expensive <S1> [mhm] </S1> [and that's] they prefer pressure sensors which are more much more cheaper , to buy </S3>
<S2> er and that , you mention in your thesis also the advantages of using pressure transducers relative to force that's from the point of calibration <S1> mhm </S1> could you could explain a little more about if we look at the cost for the calibration <S1> [mhm] </S1> [of] force sensors versus [ratio transducers] </S2>
<S1> [well well] both both sensors have to be calibrated pressure ones and er and also force torque sensors in the in the wrist but er , in er my opinion it's easier to calibrate the pressure sensor because you can calibrate it er , it er , they are easier to calibrate because there are let's say more standard tools to calibrate them with er force and torque sensors you have to use the er they are very specific of each application and you have to er find out every time a new way to to calibrate them (is) with pressure sensors they are almost er calibrated from factory and then you just have to er make some small adjustments and er with er force and torque sensors it's there are three if we are talking about er six degrees of freedom force torque sensor we have er three e- er force sensors and three torque sensors and they are a- all coupled in the same er same er mechanical packet and it's is more difficult to access and to find out a way to to calibrate them </S1>
<S2> then why (xx) as it's very easy and (say) take only very few minutes if it's well prepared <S1> [mhm] </S1> [to] make a calibration of a pressure transducer but (need a little more set) you (xx) up for calibration <S1> [yeah] </S1> [of] force sensors (so i believe) </S2>
<S3> but er once again these er these sensors if you are using a pressure sensor instead of force sensor and you pick up some load er can you weigh the mass of the load by using a pressure sensor </S3>
<S1> in theory yes </S1>
<S3> but in practice </S3>
<S1> in practice er you will have some error because er , okay it depends on the quality of the sensors of course but er for example mhm you must know not only the er pressure levels of the joint but er also distance between the er point where the load is and er the axis of rotational of the joint if it's a rotational joint and if there is for example flexibility of of the arm and you are don't take that into account you can you can get errors , it's er it's er a bit more more complex to get er right values with pressure pressure sensors </S1>
<S3> but but the (xx) force sensor attached at the end of the wrist you can </S3>
<S1> it's er almost straightforward </S1>
<S3> yes it's straightforward to get the <S1> [yeah] </S1> [mass] of the of the object , okay there are advantages and disadvantages <S1> mhm-hm </S1> related to this </S3>
<S1> and er one advantage of also for pressure sensors is that er you can er detect the forces applied at the links of a manipulator with a force torque sensor you can only detect forces applied at the end-effector and the and the tool if there is some collision er in the middle before between the ground and the end-effector you cannot find out with the force torque sensor but you can find out with pressure transducers </S1>
<S3> okay then about this this er time delay related remote remote control er in space applications you can imagine that there is a delay and it's almost significant delay <S1> [mhm] </S1> [related] to space applications but what about related to this I-T-E-R application it's it's i think it's quite short <S1> yeah bl- </S1> distance anyway </S3>
<S1> yeah in er a- if the same architecture as er i'm using it's using I-T-E-R my estimation will be that the delay can be the order of 20 30 40 milliseconds <S3>  yeah </S3> for a transmission of information another issue is if er cameras are are used then er there can be er delay also in that </S1>
<S3> er then a few words about the restrictions , you have used only single joint manipulator or would it be called a manipulator because there is one joint and one link is it a real manipulator </S3>
<S1> well you can manipulate you can apply forces or torques and er it's a very simple manipulator but er in my opinion it is a manipulator , it's missing the gripper if it had a gripper it will be would have more could be called manipulator with more weight but still you can apply forces and er move it to certain locations </S1>
<S3> yes so you you can apply forces for example if you need to get some some (xx) <S1> [mhm] </S1> [or] some surface , that's true and if you would have gripper you can lift <S1> [yeah] </S1> [simple] make simple tasks , then er page 18 er , i'm just er curious er there is a summary you are you give a summary in chapter cheve- seven <S1> yeah </S1> and then after the summary you make the final conclusions main conclusions and several topics suitable for fu- future studies , why after a summary you still , why not in the same chapter [and the main conclusions also] </S3>
<S1> [er yeah well it it could be] one chapter called summary conclusions and future studies that that could be , but er i'm using er at the end of each each chapter i'm providing a summary of the chapter so at the end of the thesis i'm providing a summary of the previous chapters i want to give the same structure , er i guess that the yeah conclusions could be in the same place as summary but er summary is a much longer than conclusions conclusions is just one page with the main findings and er you can find er more details in the summary and then even more details into the chapters so it's er it's kind of a recurrent structure you can find more details by going to the next level </S1>
<P:05>
<S2> yeah i have just a comment as er from my point of view it's a very clear structure of the thesis , the only question i was thinking about because you showed us several pictures in particular of the test facility and i'm happy i have a chance to see it in the lab but what was the argument for not showing your picture confidential or [(xx)] </S2>
<S1> [mhm no] okay in a you mean in the thesis <S2> [yes] </S2> [or in the] introduction , er in the thesis is just er i have to say that er usually pictures cannot describe well the real er device or in my opinion you might get the wrong image if you er have some not so good quality pictures not colour pictures and i i put some schematic representation of the joint because also it is intended that this is for a bit more general use not only in this application so that's why i didn't put a <SU> [yeah but i] </SU> [picture because i-] it will look at this focus only in this application and </S1>
<S2> i i accept that because this is for er scientific er doctoral dissertation er what i like very much is that you have lot of good summary in some of the main chapters that's that's very useful for the reader to have what is the highlights in the summary and , and i'm very happy also that you put in the special er section 1.4 describing the research method and your approach that's very useful to explain why </S2>
<S1> mhm-hm </S1>
<S2> <WHISPERING> (xx) </WHISPERING> </S2>
<S3> yes , okay next er chapter state of the art , er in in water , you presented moti- motivations why to use water hydraulic system instead of oil hydraulic system but there is also leakage in the water hydraulic servo valves does this leakage means that they are leaking water on the floor </S3>
<S1> no it means internal leakage er and er this er leakage in some cases is actually good because it can provide some er damping to the system </S1>
<S3> also okay that's that's true that it's internal leaking leakage , then , in this state of the art er if we think about oil hydraulic system and water hydraulic systems many of the design principles and also the solutions are are quite the same , and also on the next page page er there are for example some of the references described that <READING ALOUD> water as the pressure medium does not affect the position servo design process signify- significantly </READING ALOUD> and <READING ALOUD> main difference between oil and water hydraulics being the different characteristic- characteristics of components </READING ALOUD> and also there are several feasibility studies which describe quite the same so er is there any motivation any to do this this kind of work you have done because <S1> [well] </S1> [the] expectations are are that it's quite the same with water than with oil </S3>
<S1> yeah okay these are the expectations but er you have to show it scientifically and er also there are some things like impedance control or position control force control that er are they haven't been done with water or not so much and i think it's worth to investigate to be sure that it's actually similar than oil not only it's it's not enough that er yeah probably it's okay you have to sure </S1>
<S3> yeah i i i agree <P:05> then about this er impedance control er there are several methods described in er how to control the the forces and and positions er is it er easy or simple to describe er or to design a force controller in such , that er because you are controlling the force in one direction and position in other directions is it is it easy to make this kind or to design the controller in such a way because if you think about er arm based er manipulator there are different joints and in the joint er space it's totally different description of of the forces than in the cartesian space </S3>
<S1> well in er there is a coupling eff- a coupling effect between the the joints that er motion or forces of one joint might might affect the the other ones and er but er you're talking about the hybrid position force control where you control force in one <S3> [yeah] </S3> [direction] and position in in another , well in principle if you you can try to decouple the dynamics of the of the manipulator and try to , try to er apply force in in one direction only but er it it requires a bit of manipulation of dynamic equation of motion of a manipulator with er six degrees of freedom which is a bit complex and it can be also er time consuming for a computer system to make er online </S1>
<S3> er then in impedance control there are two types position-based and force-based impedance control and on page 22 you say that <READING ALOUD>  position-based impedance control requires accurate position control of manipulator  </READING ALOUD> but you don't say that the what about the force-based impedance control does it need an accurate force control system <S1> [yes] </S1> [because] you don't [say that] </S3>
<S1> [yes] of course also , they they consist of two er loops outer loop and inner loop and er the inner loop er is implemented the position or the force control system and er of course the more accurate it is the better the overall behaviour will be </S1>
<S3> er also you conclude that position-based scheme is preferred also in many cases instead that er if you think about electric motors they are torque controlled why is it that position-based schemes are preferred </S3>
<S1> yeah well well usually er manipulators are used for er under position control for picking something and placing it somewhere else so manufacturers already sell these manipulators with er built-in position control and er it can be that inside this position control there is this some torque control loop but er this is er at a low level access the operator doesn't have access to to that , and er if you have er if you have already an implemented and working position control er it's easier to use that one which the manufacturer has tested and has er improved than to develop your own one with er torque-based torque control </S1>
<S3> mhm-hm yeah but robots are supplied with position control loops let's and let's say which can't be modified easily it's it's quite hard work to make the modifications inside these control loops and that's the reason that the position-based are are preferred . so </S3>
<S2> yeah i i will give a comment i think all the reference you have in the state of the art is related to some very important reference from my point of view in in this area but er i will recommend to (xx) there's a lot of new research going on on intelligent control of manipulators and you you are giving a statement in your last chapter eight about conclusion and outlook so you you have pointed out this is also important area for the future but there's a lot of research going on <S1> [mhm] </S1> [and] (there's) er applications , i have sawn some- seen some (xx) (in for radio) (xx) but that's for er , i have <SS> [@@] </SS> [(xx) many (good pictures)] but some of the control (team) it could be used for (xx) oil hydraulics , on the other hand as in you have been very careful to look for robustness and that's a good argument for looking for P-I-D controllers <S1> [yeah] </S1> [that] you are you are . <WHISPERING> (xx) </WHISPERING> </S2>
<S3> okay er some words about the state of the art in human-machine interaction er usually as as you say <READING ALOUD> traditionally levers and switches have been used </READING ALOUD> and then there is er there might be a video image why is that enough because er that's the normal case of of or normal way of control manipulators and er if you are let's say in a (cage) it's it's not remote you are sitting in <S1> [yeah] </S1> [on the] chair in the <S1> [mhm] </S1> [cabin] you are looking ahead and seeing what you are <S1> [yeah] </S1> [doing] and using levers and switches and controlling and er </S3>
<S1> i think that er for example if the , one problem that you can find in the for example in an excavator that where you have levers and switches is that if you want to move the end-effector for example in a straight line you have to move several levers at the same time in the very precise way and if you have some computer system to assist you in that you just er can move er one lever which means move vertically and er the computer takes care of moving all the levers in the right er the right moment in the right place </S1>
<S3> and also the er task (xx) let's say they are quite close you can see you have let's say the humans have some stereo they can estimate the depth also <S1> [yeah] </S1> [but er] when looking at a video screen you can't estimate the depth , there is also one extra point which is missing when you are using a remote control system if you sitting in a cabin in a chair you can let's say you can feel the with your backside what's happened <S1> mhm </S1> you can listen with <S1> [yeah] </S1> [your ears] the engine how it's running and so on <S1> yeah </S1> all these are missing when you are using a remote <S1> [yeah yeah] </S1> [remote control] system <S1> and </S1> that's that's the reason you need some extra assistance </S3>
<S1> yeah and also sometimes er the smell is important when you do <S3> [oh yes] </S3> [er can] er feel some smoke something is burning then you should stop the machine </S1>
<SU> <WHISPERING> (xx) on page 26 </WHISPERING> </SU>
<S2> i have a few questions to you on page 26 , er if you have combined your schematic showing to the left the operator and then the control station and then the manipulator , er you put up a statement <READING ALOUD> the way to improve performance is to provide the operator with immediate feedback </READING ALOUD> what is that [(when you're saying that) immediate feedback] </S2>
<S1> [well that the] feedback is not coming from the telemanipulator but is generated in the control station already <S2> okay </S2> but er i can make <S2> [yes yes yes] </S2> [some] schematic </S1>
<S2> yes please </S2>
<S1 DRAWING ON BLACKBOARD, P:26>
<S1> for example this is operator control station and telemanipulator , if there is some delay between the control station and the telemanipulator knowing which are the operator's orders and if the control station has some knowledge of er the properties of this there can be some feedback can be given already here to the operator so some feedback is generated here it doesn't have to wait for this signal which is coming delayed here , <SU> [(xx)] </SU> [that would be] this is er where the immediate feedback is generated </S1>
<S2> yes thank you , then i have some questions about what kind of requirements for skills of a good operator for you have some , should you be on the top (xx) level or is it <S1> [well] </S1> [(xx)] skilled people @or@ but just (thinking) about for <S1> well </S1> application of the future </S2>
<S1> well i can er say that er in some applications in nuclear industry er where an operator is managing er some er master-arm er usually women are preferred because they are more careful than men that's one @comment@ that i can say about that , but er apart from that er it requires some training and some er some er just er experience to to control that this kind of device </S1>
<S2> now i understand when my wife is telling me she's remote controlling me sometimes <SS> @@ </SS> anyway i think we can go on </S2>
<S3> yes er one question related to page er 27 er how do you reflect the force because you are remote controlling the the system and er you lose let's say the feeling of of backside <S1> [mhm] </S1> [you] don't and also by ears <S1> mhm-hm </S1> how do you reflect the force the manipulator is er [using] </S3>
<S1> [well] well there are several several ways you can have one er numerical display that shows you er how many newtons you are applying there </S1>
<S3> how well it describes ho- if there are er number of newtons how <S1> [yeah] </S1> [do you] know what kind of forces [(xx)] </S3>
<S1> [okay this is] er one way another way will be er to show an arrow which size depends on the er force applied the amount of force applied the bigger the arrow the more force you are applying and you can also er you can also see the direction of the force er depending where the arrow is painted , for example if you have a . <DRAWING ON BLACKBOARD> manipulator in the computer screen you can have a small arrow to indicate that a force in this direction is being applied or you can have a big arrow like this one saying that the force and the manipulator is being applied here so you have a visual er visual feedback of the force that is being applied at the manipulator this is er another way to showing force </S1>
<S3> yes er this er by by using some columns or or arrows <S1> mhm </S1> that's more informative for the operator than just er <S1> [numbers] </S1> [these] numbers <S1> yeah </S1> numbers doesn't you have think about what these numbers mean <S1> [yeah] </S1> [and] what is the maximum amount of of force you can apply <S1> mhm </S1> and you don't remember maybe those ladies can remember <S1> @@ </S1> a bit better than than for example me , er what about reflecting the force by using some levers to the operator that er well he can let's say he can he can feel <S1> [mhm] </S1> [in] in some sense </S3>
<S1> well there has been studies that er they have been compare this kind of display and er this er well force feeling er feedback and er actually force feeling is er much better in terms of performance there are less errors but again we are in a trade-off between performance and cost it's very expensive to have a a system equipped with actuators and saysors- sensors and apply a force control scheme there to reflect the right force , if er you can get a similar performance with this kind of er device is much cost-effective [than than] </S1>
<S3> [yes yes] and also if you are reflecting the force to some lever er so and it's resisting your your movements so there are many people who don't who don't like <S1> yeah </S1> this kind of behaviour so i agree that this arrow or bar system or or columns they are they are much much more better and informative . [okay yeah] </S3>
<S2> [i'd like to ask] a question for 27 or the whole part of this paragraph , er i would like to here what is your view on what is needed for furtherR&D on teleoperator in human-machine interaction from your point of view and there's , you have good recording of state of the art but i think you're in new work and <S1> well </S1> maybe you have some ideas for what is most important or critical [for (xx)] </S2>
<S1> [in in my] opinion still the operator has to give quite low-level commands to the manipulator they have to say that move this joint here wait do this do that and er in my opinion it will be or when we have a more intelligent system that er even might understand spoken language you can tell the manipulator a bit higher-level orders like drop this and put it there and that's it you don't have to specify how it accomplishes the task but er it's the manipulator that decides which is the most efficient way you don't have to think every single step i think that will be the it is and will be the future </S1>
<S2> thank you </S2>
<S3> okay next er is chapter three properties of the water hydraulic drive , you have used valve-controlled er rotary actuator er , usually cylinders are used why valve-controlled rotary actuates- actuator </S3>
<S1> well with er this kind of actuator you can get er er wider range of motion than with cylinders if you think of er of a cylinder er like in er (xx) structure if you <S1 DRAWING ON BLACKBOARD, P:09> consider a cylinder actuating this link the range of motion is quite er limited with the cylinder and er with this kind of actuator you can have er up to 270 degrees </S1>
<S3> and also cylinder needs some space [(xx)] </S3>
<S1> [yeah also space this] is more more compact </S1>
<S3> yeah that's that's right , er then er also this rotating motor is it in some way comparable with with motors for example in this case i mean for a direct drive motor is this (xx) to compare er </S3>
<S1> with electric motors </S1>
<S3> yes </S3>
<S1> yeah it's er </S1>
<S3> i in the sense that er if you connect </S3>
<S1> there is no transmission <S3> [that's right yes] </S3> [between] between the actuator and the joint this er direct drive <S3> [yeah] </S3> [(xx)] you don't need any extra transmission where you can have friction or backlash or some other problems here you have a direct drive (xx) it's much much better than er than having the actuators somewhere else </S1>
<S3> er what about the movements of these er . hydraulic motors er is it er how many rotations or is it limited er [(xx)] </S3>
<S1> [this one] is limited to 270 degrees you you cannot perform continuous rotation there is some some limitations </S1>
<S3> er this you also say that <READING ALOUD> the vane actuator is equipped with a rotary pulse encoder for measurement of angular displacement and two pressure sensors </READING ALOUD> is it er in general case this kind of sensors are related to vane or is it <S1> [well] </S1> [just] in your case you have put these sensors </S3>
<S1> well the sensors are actually integrated in the body of the actuator so you can just screw the sensors there and they are very close to the pressure chambers you don't need any extra hoses or pipes to get the pressure sensors so , but er this can be applied in in general applications you you can er put force sensors quite easily to these kind of actuators in general </S1>
<S2> i have er two questions about the materials er i i know you have er shown it to me that earlier there have been er first there've been prototype for the test facility using stainless steel and some polymers or (xx) for all the seams and in the newest test facility they are using titanium and it's also mentioned in the appendix , what is the main argument for er using titanium </S2>
<S1> er it's er lightweight </S1>
<S2> lightweight <S1> mhm </S1> so you have a lower mass [(when you use it)] </S2>
<S1> [yeah] and er it's er still water compatible material </S1>
<S2> then i have a question about the servo valves er you have stated in the table A two on page er 133 the natural frequency (approximately) hundred hertz and the damping ratio of point five <S1> mhm </S1> er my question is because i'm looking for any measurement using water for having frequency response shown as a (xx) at my lab we have some ultra <S1> mhm </S1> servo valves working running through water but the (xx) we have got is based on oil <S1> [yeah] </S1> [because] it's (a lot stable) then i've been asking is there any measurement er (xx) in water </S2>
<S1> well actually those those figures are from the manufacturer's data sheet <S2> yes </S2> they are not from measurements </S1>
<S2> but if we compare to if we have what we call in er servo valve terms hundred per cent rated input <S1> mhm </S1> then we can identify the natural frequency in the damping ratio <S1> yeah </S1> but if we only have say forty or fifty per cent input will it be faster [or slower] </S2>
<S1> [yeah it will be] will be [faster] </S1>
<S2> [yes yes] thank you , i have one more question that's about i know it's schematic for the practical er after (xx) necessary to have some accumulator in the system because you have a problem with fixed displacement </S2>
<S1> mhm er if i remember right the we didn't use an accumulator in this case . but er we , we had er quite er constant er flow and pressure level so we didn't find it necessary </S1>
<S2> <WHISPERING> (xx) </WHISPERING> </S2>
<S3> yes on er page 30 about these models what is your contribution in in deriving these these models <S1> well [i] </S1> [(xx)] models </S3>
<S1> i'm a user of existing models i understand how they work but i haven't created them and i'm just using them </S1>
<S3> what about the parameters </S3>
<S1> parameters are mainly found experimentally and er experimenters experiments have been designed to find out the parameters that fit in those models </S1>
<S3> have you done these experiments or </S3>
<S1> most of them </S1>
<S3> most of them so you have taken these models and find out these these <S1> [yeah] </S1> [parameters] for these these models </S3>
<S2> i have a question er you have stated it clearly we have a nonlinear model and a linear model so nowadays we have a strong simulation software like matlab simulink or similar so why do we not only use nonlinear models and because there's no need for the linear model i know you have given some <S1> [yeah yeah] </S1> [statement but] i'd like to hear your arguments again </S2>
<S1> with er with nonlinear models what you can do is to test test and test and see what happens it's like a trial and error , with linear models you can find out the right value if you think a bit faster than with nonlinear models , if you know that er in order to guarantee stability your gain has to be below this number er after that you can tune you for example your nonlinear model more easily than if you start testing values in nonlinear model , and er you get a better insight of er what should be changed or what could be changed er to improve performance you you get er you get a more in-depth view with a linear model of the dynamics of the system , usually the dynamics of a system might be a bit hidden in er nonlinear models </S1>
<S2> i agree but you stated it somewhere so it's very useful for <S1> [controller design] </S1> [controller design] and then when we have to design the controller then we have to , make [all the simulations (xx)] </S2>
<S1> [yeah finetunes of parameter] </S1>
<S2> okay . <WHISPERING> (xx) </WHISPERING> </S2>
<S3> yes er page 32 verification of the models , there's this nonlinear model you have verified and also the (xx) in figure five simulates simulated and experimental response of of the vane actuator , er can you say anything about the figure five because on on the same page figure s- six you have zoomed out the let's say the main behaviour and i like this this way of zooming out because in my opinion from figure five you can't say anything <S1> yeah </S1> almost anything about the beha- behaviour <S1> [mhm] </S1> [of the] model how how well it behaves <S1> mhm </S1> because diminishing the the picture you can say that it it works very well </S3>
<S1> yeah yeah the more you @reduce@ the scale the better it fits , but i also have thought of of putting some error plot but in my opinion it was more clear with this zoom er zoom plot </S1>
<S3> why not er on the next page you on figures figures eight seven you also present er the linear simulation model why did you zoom out </S3>
<S1> well the differences i think are already clear in the not zoom plot , <S3> [yeah] </S3> [and there are] there isn't any need to zoom because er you can already see the difference you can already see that there are some minor errors </S1>
<S3> yes er in in in fact er the linear model has errors in in damping <S1>  yeah </S1> because the the velocity is behaving a little different with the model than with the measure measured value , but er you also stated that er after you have verified let's say this er nonlinear model which behaves quite accurately compared to this er linear model which i would i would say doesn't behave so accurately especially damping is not accurate <S1> mhm </S1> but you state that <READING ALOUD> in general the simulated response with the linear model agrees well with he nonlinear model and with the measured response and therefore the linear model can be used for controller design </READING ALOUD> </S3>
<S1> well </S1>
<S3> is that really so </S3>
<S1> at least in steady-state values of position and natural frequency er yes , and damping does not agree in the er valve opening but it does agree in the valve closing so i think it's quite close to this is i would say as close as you get with linear models </S1>
<WHISPERING, P:19>
<S3> okay i would like to ask on page er 35 you compare , that your you say that er <READING ALOUD> the only parameter that varies with the pressure medium is the effective bulk modulus </READING ALOUD> er and the bulk modulus depends on the , <READING ALOUD> the effective bulk modulus depends on the bulk modulus of the fluid on on the flexibility of the pipes and hoses </READING ALOUD> er what about the can same pipes and hoses be used in water and even oil hydraulic cases </S3>
<S1> mhm yes as long as materials are compatible , in this case er we didn't have er either pipes nor hoses the connection between the valve and the actuator was direct in the actuator body </S1>
<P:08>
<S2> i have a question , you you are talking also about internal leakage between the actuator chambers (with) water versus er oil , er the seals i have identified you have in this vane rotary that is very important for controlling the internal leakage </S2>
<S1> yeah , and er let's say that in principle internal leakage is a bad thing but er it it can add damping to the system which is a positive positive er effect , and er also frictions might affect also friction er between the vane and the er outer body and er this has to be careful with fri- er sealing selection to to have a low friction sealings </S1>
<S3> okay on on page 38 er you derive er steady-state position error equation 14 and er <READING ALOUD> the position error can be reduced by decreasing the effective leakage and by increasing the controller gain </READING ALOUD> controller gain er how high this gain can be increased can the error </S3>
<S1> well [the] </S1>
<S3> [can it] be reduced can it be reduced to zero by increasing the <S1> [er no] </S1> [gain is there] any limitations </S3>
<S1> yeah in the position control system there is a limitation for the er position er proportional control if , er stability is desired and this limit is in in equation 12 which er er establish which is the value that er proportionally has to be in order to guarantee stability so you can reduce position error till certain limit but er not to zero </S1>
<S3> can is it the error about the same because er on in water hydraulic case you can use higher gain than in oil hydraulic case <S1> yeah </S1> what about er steady-state or is it about the same amount in both cases because it you can use high gain in water </S3>
<S1> mhm yes but as as we see the equation 14 the proportional error depends on natural frequency in equation 12 and er natural frequency of water is about 1.2 times bigger than with oil so you can use proportional gains 1.2 times bigger than oil , but on the other hand the leakage co-efficient is er ten times er ten times er bigger with water than with oil so you have to you will actually have more error with water than with oil if you check this ratio between leakage co-efficiency and er unproportional gains , you you cannot increase the proportional gain so much to compensate the ten times bigger leakage of of water </S1>
<S3> okay but this was very very good explanation but er it wasn't printed on on the on your <S1> mhm-hm </S1> not so clearly </S3>
<S2> i have a short question er you stated on page 38 <READING ALOUD>  other types of linear controllers such as P-I P-I-D controller </READING ALOUD> and that (will add an integration) but if we look at the position controller we already have an integration from the velocity (aim) velocity to the position <S1> yeah </S1> have you been thinking about if you could use an approximation to a P-D controller a leak controller </S2>
<S1> er i have to say that er i i thought it but er i just wanted to keep it a proportional control only if if possible and er even proportional derivative control even if it's still simple but er proportional is even simpler and er i wanted to keep things simple </S1>
<S2> is it the other argument we have talked about it earlier in this er defence that's also you could have more robust systems <S1> [yeah] </S1> [with] P-I-D controllers . <WHISPERING> (xx) you continue </WHISPERING> </S2>
<S3> yes , can turn lot of pages , well stop on page er 42 just to say that at the beginning i criticised the thickness of the thesis too many pages but er this also results in some good aspects i would say that derivation of equation is is clear and indicated with explanation and it's a lot of easy to follow and it is it's very good point in your in your thesis , on page er 43 you mention the stiffness some ways , yes <READING ALOUD> let us now consider that the actuator is moving slowly and that it is in contact with a stiff environment </READING ALOUD> what kind of environment is a stiff environment </S3>
<S1> well here this stiffness environment is modelled as a torsional spring and it means that there is some compliance when er being in contact with the environment , for example you could consider that this table when you press it down it er goes inside a bit and goes outside a a bit so this er the spring constant of this table will be very high you can model this environment er with with a spring </S1>
<S3> with a spring with high </S3>
<S1> very high er stiffness </S1>
<S3> yes <P:22> er on page 49 you mention that er , er er this backdrivable what do you mean by backdrivable </S3>
<S1> that er you can er a force exerted on a manipulator will cause a a motion </S1>
<DISC CHANGE>
<S2> i i have a comment , i appreciate that you in figure 13 also have used the root locus as a part of analysis for stability and the dynamics it's it's <S1> mhm </S1> introduces torque (xx) , i also appreciate on the next page er 48 that you also give a good statement for why will it help to add the P-I controller to introduce an integrator for the steady-state torque error , er so that's a good contribution that you take the advantage of (xx) controller for that also </S2>
<S1> yeah </S1>
<P:09>
<S3> er yes in the summary of in this chapter three you introduce this nonlinear model but you didn't use it you just <S1> yeah </S1> did everything by by using this linear [model] </S3>
<S1> [yeah] , well since er i'm describing the characteristics of the of the drive and er i'm giving the linear model and verification er i think my opinion is to put er all the models that are used in this thesis together so you can compare for example that er plot of er behaviour of dynamic er er linear model and nonlinear model and experiments and you can compare both very quickly , and er it is later when the nonlinear model is used but in my opinion it belongs also in this chapter because there are basic characteristics of the of the vane actuator can be already seen with the nonlinear model </S1>
<WHISPERING, P:49>
<S3> er just to s- some questions er , page er 59 you have described the block diagram of the position-based impedance control scheme with with pressure feedback and on on figure 17 , is er i wonder what's the difference between these two two diagrams these two [figures] </S3>
<S1> [well] in figure 17 in page 57 the feedback the measure contact torque is er coming directly from the environment block <S3> mhm </S3> while in figure 18 in page 59 it's the actuator torque which is going to the impedance model so , i- actuator torque or er contact torque is introduced in the impedance model </S1>
<S3> and the sensors are also different </S3>
<S1> yes with er in page 57 the contact torque is er measured with a force sensor or force torque sensor and er , in page 59 actuator torque is measured with er pressure transducers </S1>
<S3> yes then on the same page er you say that er let's read it from here <READING ALOUD> the hydraulic manipulator will behave with the desired impedance only if the position control system is able to follow the modified trajectory accurately at every moment the position-based impedance control scheme depends on the performance of the position control loop </READING ALOUD> and then you say <READING ALOUD> in this thesis a proportional controller is used in the inner position control loop resulting in a system which is not optimised for position tracking accuracy </READING ALOUD> is there any contradiction between these arguments , you demand an accurate position control loop <S1> yeah </S1> and then you use , er a system which is not optimised for the [position] </S3>
<S1> [yeah] well what i want to say is that er there has to be a position control in er position-based impedance control and the implementation of this position control is almost free you can use any er er position control scheme that you wish like er some nonlinear or P-I or P-D , but er i'm just er using one which er is allows to have a simple linear model and you can analyse er the effect of different factors but er the important thing is that this is this is not limited to er proportional impe- er proportional control you can use any any er position control </S1>
<S3> okay and er once more <READING ALOUD> however the use of proportional controller enables the carrying out a linear analysis to study the dynamics of the system </READING ALOUD> so you have made also some simplifications </S3>
<S1> mhm well i have chosen a simple controller good enough but er simple enough that allows to see the effects of of different parameters in the performance of er of the system er , er better behaviour will be obtained with er for example nonlinear control that compensates nonlinearities but er then it wouldn't be possible to er analyse what happens if this value increases and this value decreases it could be done with er nonlinear simulation but er not for control design </S1>
<S3> er on the last sentence you say <READING ALOUD> any position controller that compensates nonline- nonlinearities and accurately follows a reference trajectory can be used </READING ALOUD> er can you say any any limit for the accuracy it's that's the word what do you mean what what kind of accuracy is needed </S3>
<S1> well actually accuracy depends on what task you are carrying out of course the better accuracy the the better it will behave , but er depending on which task you're carrying out perhaps is not necessary to have er high accuracy for example if you , manipulator wants to er grab some object if er if the surface of the gripper are prepared to er handle this cylinder shape you don't need the exact location it will er accommodate this this object more easily , so it depends a bit in the in the task that er needs to be carried out , obviously the better the , the controller the better will be the performance but er again it's a matter of trade-off between performance and complexity and robustness </S1>
<S2> yeah i agree with my colleague first be careful when you just put it in this way accurately and @(xx)@ 'cause how accurate in <S1> [yeah] </S1> [(xx)] but you are of course aware of , and i have a question to the same page you have it just er below equations 72 <READING ALOUD> in order to obtain the desired behaviour of the manipulator it is necessary to know the inertial load </READING ALOUD> that's not so difficult <READING ALOUD> and the viscous friction </READING ALOUD> so how will you comment on </S2>
<S1> well er actually this comes from , er this is in the case that you are using pressure sensors <S2> mhm </S2> to estimate the contact torque and er for example if we write the dynamic equation of motion of er of the manipulator is something like this <WRITING ON BLACKBOARD, P:07> where er this is inertia and viscous <S2> [yes] </S2> [term] so if we want to know this we want to estimate the contact er from actuator measurements with pressure we need to know this this and this , and er well er you can carry out some simple measurements some er some friction er measu- friction identification measurements to identify this viscous friction term </S1>
<S2> and sometimes it's also possible to make in the lab an experiment because if you identify the viscous friction you can look for the damping ratio given the inputs <S1> mhm </S1> (response) thank you <WHISPERING> (xx) </WHISPERING> </S2>
<S3> we can continue even in the page sixty- 62 . er on the last sentence you say er not on the whole page the last sentence on the whole page but er chapter 4.1.3 <READING ALOUD> however the position-based scheme is superior to the torque-based scheme for practical imple- implementation reasons and therefore the rest of this thesis is focused on position-based impedance control  </READING ALOUD> so you are in favour of the position-based scheme er for implementation reasons what about performance reasons w- which one is important </S3>
<S1> well let's say that er in case of hydraulic actuators . mhm force control or pressure control hydraulic actuator is a bit more complex you will be you will get actually er better performance because the er pressure force loop is faster than the position one , but er these advantages er will be er destroyed by using er or by or the disadvantages like amplification of measurement noise and and so on , but er in if we are dealing with ideal components it will be better to use force-based but er we don't have ideal components </S1>
<S3> (fine) yeah then chapter 4.2.1 types of environments er , you er <READING ALOUD> the different types of environment can be expressed as a combination of inertial resistive and capacitive environments </READING ALOUD> how do you describe for example the city environment or the rural envir- environment </S3>
<S1> or the </S1>
<S3> city environment and rural environment can they be described by this kind of </S3>
<S1> well i i mean er for example when you're carrying out a task er you are in a quite er known location with er you can model each objects each object as , when you are modelling you are not er you are capturing the main properties of the object not all the the properties and er , with inertial resistive and capacitive it's a way of simplifying a er the real world into a world understandable by er mathematics or physics </S1>
<S3> and also by the environment in this case you mean you don't mean the whole city environment [you just (xx)] </S3>
<S1> [no in the in the (work) area] </S1>
<S3> yeah that's right that's the basic <S1> [okay] </S1> [main] difference . er then er you describe this er some places different environments they are on this page 63 what kind of environment types are inertial resistive and and capacitive capacitive environments , er can you give any examples of what kind of when you use for example resistive environment [type and] </S3>
<S1> [yeah] in in resistive environments the viscous friction is er dominant so you can think for example in underwater manipulation the effect of water is er it causes some er viscous friction on the motion of the manipulator so that would be an example of a resistive environment , and er inertial environment can be can be compared as a manipulator holding er er big load or that er increases the overall inertia of the manipulator , and er with capacity there is a spring effect as in the table for example </S1>
<P:14>
<S3> er on page 65 you introduce er Q with er subscript E small E what does this mean it's not mentioned here , what does this er notification describe </S3>
<S1> which which one excuse me </S1>
<S3> er page sixty- 65 and on the equation eighty- 89 there is this <S1> [Q-E] </S1> [Q] with subscript E </S3>
<S1> it means <S3> [(xx)] </S3> [location of the] location of the environment </S1>
<S3> yeah but where is it mentioned what does it describe </S3>
<S1> mhm-hm </S1>
<S3> tell me if if you find </S3>
<S1> well no i didn't well it's it's missing it's location of the environment </S1>
<S3> yes it's quite obvious if you read a little bit more that it describes the the position of the environment er but it's not mentioned even in the nomenclature <S1> yeah </S1> , is it known , how </S3>
<S1> it will be good to know it but it is not always known in the , if you want to handle for example this glass <GRABS A GLASS ON THE TABLE> er this glass could be here or here and er you can er depending on er , on the properties of the manipulator if it's in the wrong location it could er break it or not even grab the glass so it's it's important to know it but er if you don't know it er impedance control is such a controller that er can minimise on contact forces </S1>
<S3> er related to this er same topic let's say this equation er 91 on page 66 there is this er er contact torque T-E equals er K-E times Q minus Q-E and if you don't know Q-E what is that </S3>
<S1> well er this is er contact torque contact torque so you can measure it with force sensor or estimate it er with pressure sensors </S1>
<S3> so you can let's say [indirectly] </S3>
<S1> [yeah] you can estimate it if you know the other parameters </S1>
<S3> er is it the same on page 68 er equation 100 er there is this position error Q ref minus Q-E </S3>
<S1> yeah it's er </S1>
<S3> if you don't or do you know this er external- er environment position </S3>
<S1> mhm no but er the error that you obtain er this is at a theoretical level the the error er will be will tend to this difference , if er you don't know it you can estimate it and perhaps er (bound) the er error and er obtain some desired behaviour </S1>
<S2 AND S3 WHISPERING>
<S2> er we move forward to chapter 4.3 page 75 , could you briefly explain a little more about the box you have here in the block diagram in figure 24 you have first (given) the name nonlinear model and what's inside the box </S2>
<S1> well there is [er] </S1>
<S2> [could] you just explain [briefly] </S2>
<S1> [there] is er nonlinear model of a servo valve there is a nonlinear model of the actuator which includes er internal leakage and er nonlinear friction model and there is a model of the er mechanism which is the inertial mass , and er these models are connected er together in order to provide the overall er simulation model of the v- vane actuator </S1>
<S2> okay </S2>
<P:05>
<S3> page 80 er about this er static torque control , you des- design an overdamped and you got an underdamped system <S1> mhm-hm </S1> can you explain why this [happened] </S3>
<S1> [well] the the combination of the actuator with the the environment results in a new er dynamic system expressed by equation 117 so er using these er values of new values of er er desired or stiffness and er all values of er damping and er inertia you get a a lo- er different damping if you use . you use er equation 115 in page 76 with er eq- with er the new stiffness combined stiffness of er , of er the manipulator and the environment you get a a different damping </S1>
<S3> so on next page you are using different damping ratios er <S1> mhm-hm  </S1> you are using damping ratio two and damping ratio six , er to get a damped system is six enough or because two doesn't seem to be enough </S3>
<S1> no [well] </S1>
<S3> [how do] you how do you say what is enough what kind of damping ratio is [enough] </S3>
<S1> [well] this is a bit related with with what was mentioned before of er how much do we know about the environment if we know what's the stiffness of the environment we can accurately calculate which is the er needed er damping term , but er if we don't know in order to be on the safe side we should use a large damping ratio , large well six it's enough for er quite a stiff environment of er one million newton metres per radian </S1>
<S2> i i'd like to continue (following you) as er in figure 29 you have shown the two cases as er when first the er starting on the (xx) will the damping er for 2.0 and 6.0 and you put a statement and i understand that to avoid (xx) in contact (situation) because you don't like damage er <S1> [yeah] </S1> [anything] so have you if you should give an advice would it also be enough (were) the damping er say 4.0 or 5.0 or is it to be a robust system so you say to assure you </S2>
<S1> yeah yeah [this is a] </S1>
<S2> [(xx)] 6.0 </S2>
<S1> yeah for example you can use a nonlinear simulation with an estimation of the er environment stiffness to find out this er right er damping values that's that could be one use of a nonlinear model </S1>
<S3> then we move to chapter five , this operator telemanipulator interaction and control station , er you talk about talk about full automatic control and also manual control are there any er in between of these er just to jump from a totally manual system to a totally automatic [system (there is)] </S3>
<S1> [yeah there are] several levels in the middle like er share control where the operator er takes pla- takes care of some actions and the computer system takes care of some other actions , for example what we were talking before that er the operator indicates some trajectory for example vertical trajectory for the manipulator and the computer system calculates which is the er joint rate that you have to apply , on the other extreme will be manual control where the operator will be with levers calculating by himself or herself the right er combination </S1>
<P:09>
<S3> then er , chapter 5.1 operator control station interface , <READING ALOUD> the computer features a virtual world containing two 3D virtual models of the telemanipulator for issuing control commands and for feedback visualisation termed master model and feedback model </READING ALOUD> and this reference er 54 is is your reference so does it mean that it's it's your idea you have introduced the idea of of describing [these (xx)] </S3>
<S1> [well i] introduce the notation of using master model and feedback model some other authors use er master manipulator the slave manipulator or feedback manipulator i just used some er nomenclature that er it er it sound er sufficient for me that er it er gave a description of what was really happening </S1>
<S3> er then er <READING ALOUD> the operator moves the master model by changing individual joint values or by selecting the desired location of the end-effector using keyboard and mouse </READING ALOUD> how user friendly user interface this kind of is you just use keyboard and mouse </S3>
<S1> with er keyboard is not very user friendly because you have to go to some textbox and then type numerical value but er with er mouse you can select one frame and tell the manipulator to go to the selected frame so it's it's quite intuitive you don't need to know any location you don't have to write any numbers you just tell with a mouse go here and that's it </S1>
<S3> is it designed for a scientific user what about an ordinary user who just jump in and you say that try to control this kind of system and say that here is the mouse and here is the key- keyboard </S3>
<S1> well i haven't tried with a non-scientific user , i would say that is simple to use but er , well i haven't tested the and it should be tested perhaps to er to evaluate the usability of of this device </S1>
<S2> i would like to ask you er on page 90 , i think it makes sense at this stage for the research and to design and implemented the software , (but you put on) the statement er <READING ALOUD> the operator moves the master model </READING ALOUD> and then you focus on using keyboard and mouse controls if you look for the future what will you think about for the interface as a this is a classical interface <S1> [yeah] </S1> [there's] a keyboard and </S2>
<S1> well actually one er simple step is to use some tactile screen that you can point where you want to to go , next step would be to have er perhaps a three dimensional display you wear some special glasses perhaps and you can er perhaps even grab the manipulator with your hand and bring it to the desired location , i think that's the future or or not the future because it's already here but er that's the development of these techniques will be very important in the future </S1>
<S2> okay , <WHISPERING> (xx) </WHISPERING> </S2>
<S3> yes go ahead to pag- page er 92 you are talking there <READING ALOUD> roundtrip transmission delays are about point four seconds  </READING ALOUD> so delay is er less than half a second , er then you are wal- talking about move and wait strategy , well if the delay is less than one second how much in- er human can wait i don't think that he can even sense that he's waiting </S3>
<S1> no probably not then it becomes a move and watch strategy <S3> yes </S3> when the delay is close to zero there is no delay at all </S1>
<S3> er on page er 93 you <READING ALOUD> the main performance concerns are reliability and robustness </READING ALOUD> and and so on so you concern about the reliability and then er on the same page you say that <READING ALOUD>  communication link between the control station and the water hydraulic telemanipulator is carried out using the T-C-P I-P communication </READING ALOUD> and you say that er <READING ALOUD> communication protocol because of its high bandwidth and reliability </READING ALOUD> <P:14> okay yeah okay it was a little bit wrong , talking thinking about er </S3>
<S1> well there are several methods to <S3> yeah </S3> there are many many methods to transmit information but er T-C-P I-P is quite common meaning that there is a lot of documentation it's easy to implement and er then it er it has a very robust and reliable , and er for example we can think that er internet is using this kind of protocol and er we can think that sometimes we lose emails or we cannot access web pages but we can use this protocol in some close network where no-one has access there so you can still you the same protocol in only for connecting two computers </S1>
<S3> mhm and then er you are using an event-driven system when op- er (xx) a security mechanism is also included what about if no event is happening what happens with does the security system react </S3>
<S1> yeah or there are like er two events er one of receiving messages and er another event which is checking if er a message has been received and these two things are are er running simultaneously and er once er message is received then it er it tells to the other event or the timer handling event that okay i received a message so you can start the timer again and er like this it's working , if no message is received the timer will increase till certain value i think it's about 200 milliseconds if er a value hasn't been receiving 200 milliseconds the er timer event will er jump and er will stop the system </S1>
<S3> mhm-hm , then er on the next page you are talking about a trusted source source that you check that the package is coming from er a trusted source , er it might be the source might be okay but er you don't know who is using this tru- trusted source you don't check the person who is using </S3>
<S1> no if we think in modern computer systems you can er put a password to the computer so you can know that this is an authorised user </S1>
<S3> yeah <P:08> okay page er 95 , you say that er <READING ALOUD> the tests carried out show that the user interface is able to graphically show the relevant information to the operator in an intuitive and easy-to-understand way </READING ALOUD> how you have checked that it's easy and easy to understand and intuitive <S1> [well] </S1> [it it] might be intuitive but what <S1> [yeah] </S1> [about] the e- easy to [understand] </S3>
<S1> [but] this is similar when we were talking before about having some numbers showing forces and positions or having some er arrows or some er position display er in a 3D model , the the operator can can identify where the manipulator is just by looking there watching at it and not by reading er ten different joint position numbers is much easier to to look at er some three dimensional model than than to a set of numbers </S1>
<S3> but usually if if you make this kind of statement it means that you have performed usability tests so you have selected a certain amount of persons and and made the tests and analysed the results and then you can make this kind of statement </S3>
<S1> er it it wasn't er done with such a er methodical way it was a more er asking er was it easy to use yes and that's it , i didn't measure any task completion times or something like that it was a more er subjective opinion than than er with numbers </S1>
<S3> so asking your research fellows <S1> mhm </S1> how do they [(xx)] </S3>
<S1> [and of course] er they are from scientific <S3> yeah </S3> [community] </S1>
<S3> [yeah they they are] okay they are thinking as you are thinking and <S1> [maybe yeah] </S1> [that's] an expected result </S3>
<S2> but i have a question er to you you use a term on page 95 er the statement is <READING ALOUD> by using the control station with the dynamic simulation model the operator can investigate interactions with the environment and get a good grasp </READING ALOUD> what is a good grasp </S2>
<S1> can er </S1>
<S2> could you explain what you [think (xx)] </S2>
<S1> [can i] get in in a second how the manipulator is working er meaning that er if er , if the system is overdamped you just er try it with with a system and you will see that <FOREIGN> oho </FOREIGN> this is going this way and then you immediately see that er what's the dynamics of the system , but er a good grasp means that er immediately in er it shocks you that it's like that </S1>
<S2> okay , would you like to continue </S2>
<S3> okay er then on page 96 water hydraulic manipulator , er <READING ALOUD> the roundtrip delay when the water hydraulic manipulator is used is connected to the control station is of an order point five milliseconds </READING ALOUD> and i remember that er distance were something 30 metres <S1> mhm-hm </S1> in your case so the cable was as you say the communication lines . so the distance is quite short and in that is the reason , okay there the <READING ALOUD> there is no significant tis- time delay in this implementation of a teleoperate- operation system because the distance between the remote and the local site is relatively short , the bandwidth of the communication lines is high the computer power in the data processing unit is kept to a minimum and the inertia of the manipulator is not not high </READING ALOUD> so in your case the distance is so so short i remember s- <S1> [er 30 metres] </S1> [somewhere it was] 30 metres and the gu- cables were only 30 metres what about if you take a big roll of cables and go several times around the university to get a little bit longer distance <S1> [mhm] </S1> [to] get a little bit longer delay </S3>
<S1> it will get longer definitively but er but for example with er with such a short distance and er such a high transmission bandwidth er there are some other limitations er i mean that er you cannot directly relate er velocity of transmission distance and er and er delay because er there are some other operations that er take time like er every time you send a message you have to wait a bit and every time you receive a message you have to wait a bit plus the velocity the transmission rate in the cable so you cannot directly calculate the the delay knowing the bandwidth and knowing the distance er there are other factors in both </S1>
<P:21>
<S3> er about these tables three four and and so on , how do you interpret these these tables how do you find out er you have it was not so easy to me to follow these tables <S1> [mhm-hm] </S1> [what] you are trying to say regarding these , you are using graphical inter update rate <S1> mhm-hm </S1> low or high and and so on and also control station and si- simulation server and <S1> [mhm-hm] </S1> [control] station and tele- telemanipulator , could you er explain [what] </S3>
<S1> [yeah] well i i tried to analyse how different factors er affect the the delay if er if you have a very complex virtual world with a lot of objects then that this requires certain computer power to manage all these graphics if there are a lot of graphics the computer processor is er dealing with the graphics and is not transmitting information because there is the share of the processor load is shared between the er transmission of information and display of images , so if we have a a low graphical update if we are just er showing the virtual world but we are not moving around all the time then the the computer that doesn't have to draw all the time the the virtual world and there is more time for transmitting information so there is less delay , also if you are using a simulation model to get immediate feedback if you have a nonlinear model this requires more calculations than a linear model that's that means that er the computer is er dedicating more time to simulating than to transmitting information so the delay is er increased if we use a nonlinear model because computer cannot handle both things at the same time </S1>
<S1 AND S2 WHISPERING, P:34>
<S2> yeah i think if we look at some of your shown figures where you have measured position in er figure 39 you have shown some different cases with the damping from er 1.0 to one five . and that's with the pressure transducers and when you mention pressure is it you measure the absolute pressure is that so that (phi) one and (phi) two or is it (delta phi) </S2>
<S1> er there are two pressure transducers <S2> [yes] </S2> [and] i measured both pressures and <S2> yes </S2> then i derived the actuator torque by er the difference of pressure multiplied by the volumetric displacement </S1>
<S2> and that's how i [understand it yes] </S2>
<S1> [and with that] it's obtained the actuator torque and from that is derived the contact torque </S1>
<P:10>
<S2> and and you have also (looked for) what you call low inertia and high inertial load that's in figure four- . one and it don't look like to be so sensitive to er the change from low inertia to [to high inertia] </S2>
<S1> [well i'm i'm] using slightly different controllers for low inertia and high inertia <S2> i (thought so) </S2> yeah with er in case of the high inertial load i'm er the effect of the higher load will cause some er some er position errors due to the effect disturbance effect of the of the load , but er i compensate those errors by by giving some er (fit) forward command to the to the (xx) , and er also of course there is some gravity compensation and some basic er friction compensation </S1>
<S2> so er the settling time is very similar if we compare low inertial load with high load </S2>
<S1> that means that er they are behaving according to the desired impedance </S1>
<S1 AND S2 WHISPERING, P:11>
<S2> yes on page 106 </S2>
<S3> 116 analy- this chapter analyses the results , you say that er <READING ALOUD> the position errors observed during the experimental tests are different to those in the nonlinear model due to the different reference command signal </READING ALOUD> why you use different command signal in in simulation than in [experiments] </S3>
<S1> [well in er] in simulation i was using step input of 0.1 radians and in er in experimental measurements i was using one radian , well er i was using 0.1 radians in simulation firstly to first to be able to compare with some other authors which were using also 0.1 radians and er and er also because er , with the nonlinear simulation model er i tried to tune a bit the the parameters to er find out er some right value for example control parameters and er also with er the with nonlinear simulation er the motions are quite around the zero position er where there is no er there is no load applied to to that actuator that means that er there isn't any <SIC> disturbant </SIC> effect by by the load so it's more easier to to check if everything is working as it should and er then s- then once i know that these control values are okay and the system is behaving as it should then i increase the the step er reference signal and and then i use it in measure- measurements </S1>
<S3> yeah okay but er you don't you didn't have time to do same (xx) as as in , just to test [the behaviour] </S3>
<S1> [er] , i had i had time but er again if i had put er if i had made the same test with 0.1 radians and one radian er i will have too many redundant figures showing still the same , i think it's easier to to say with er one sentence that er the nonlinear model and the experimental results are on the same level because the same response is obtained not the same numerical response but the same behaviour is obtained </S1>
<S3> and once again more pages </S3>
<S1> more pages yeah </S1>
<S2> i have a comment to appendix , in table A five from my point of view it makes sense the high quality you have been using of force sensors er bofors elektronik with a high accuracy plus minus one point (three on full scale) and similar for the pressure transducer and also the heidenhein position encoder so that's very good you have put that in your thesis <S1> [mhm] </S1> [so] for other teams when you say you have used the professional equipment for the results , but i have a question on figure 43 if we look at the graph showing with the damping ratio 2.0 in the beginning it looks like there's some oscillation or what kind of problems or phenomena is shown in this part of the graph , do you have some comments to that </S2>
<S1> yes er i have to say that er i don't have a clear answer for that er i would say that er this is caused with some nonlinearities of the valve for example hysteresis , and er also might be also caused by some internal leakage and er i think er it's worthy to to focus in future to make some er further research to study this phenomenon , also it can be because of the nonlinear behaviour of er friction can be some stick-slip er friction effect but er i have to admit that er more research in the future is needed to to be able to find out the cause of that and find out a solution </S1>
<S2> then if we could move forward to page 131 figure 44 you show measured torques two cases again for damping 6.0 measured with force sensor and the other case pressure transducers in the control loop , again if we look in the beginning er from zero to one second there's a little set-off between the two graphs </S2>
<S1> yeah that means that er gravity is not correctly compensated for pressure sensors </S1>
<S2> so that's the explanation [for (xx)] </S2>
<S1> [yeah there] is some er error yeah </S1>
<P:18>
<S2> please </S2>
<S3> okay the final question from my point of view is related to appendix B on page 136 , er you mentioned er related to this force reflection to how to sense the forces you mention these arrows <S1> mhm-hm </S1> just written also on the paper still seeing there , where do you found these arrows because if you look at these figures you don't find </S3>
<S1> yeah [you you can] </S1>
<S3> [anything] related to </S3>
<S1> you can see them in page 140 in figure B four </S1>
<S3> B four </S3>
<S1> yeah , do you see in the contact place between the manipulator and the task or the environment there is an arrow which gives the direction of the force and the size of the force </S1>
<S3> yeah because this is er not a colour picture so </S3>
<S1> yeah perhaps it's [not so clear] </S1>
<S3> [it's not very] easy to seeing that there are these these forces okay that's that's fine , thank you </S3>
<S2> er i have a question on page 121 there's a section called the summary and also in page 122 , i'll take the first part of it (xx) put it up here , on page 120 <READING ALOUD> the experimental results indicate a relative position error of point one per cent resulting in a position accuracy one order of magnitude better than in the case of an oil hydraulic manipulator </READING ALOUD> is it for similar as as for similar [(xx)] </S2>
<S1> [well it's] </S1>
<S2> or set-up or </S2>
<S1> i'm giving a relative position [which] </S1>
<S2> [you've] stated it very general </S2>
<S1> yeah and er i have to say that it's it's very difficult to to compare with other authors because it's not always given all the experimental conditions and all the characteristics of the manipulator and the environment and er there are many articles about impedance control with electric manipulators and even hydraulic ones but er there is always some er er data missing in order to calculate all the position or torque force errors and er this is one estimation with er with one oil hydraulic manipulator the the same er , okay er relative position error means er er reference position minus er er measure position divided by reference position so it's er a scale to the er step size that you are giving , so it is given in percentage it gives an an idea of er that you can use to compare different manipulators </S1>
<S2> so er my last question and then i have after that just a short comment , er on page 121 er have you been thinking about using (adapt to) learning controllers or learning controllers and you put up as a point for research in the future on page 126 <READING ALOUD> adaptive set of impedance parameters for contact and non-contact situations </READING ALOUD> </S2>
<S1> mhm-hm for example er as i explain in the chapter relative to impedance control when we are controlling the manipulator in motion in free space it's important that the desired stiffness is high in order to have er small position error , but on the other side when you are applying some force to an environment it is desired that the that the desired stiffness is as low as possible so there are these two contradictory er ideas , so in the future it will be interesting to to change the valuenof the desired stiffness depending on if we are in contact situation or in free motion and er at first this change would be quite radical from er (ratio) for example hundred times this value but er in the future it could be some smooth transition between er free er values with free motion and values with with contact , i think er it's it's worth er researching that that area </S1>
<S2> thank you i have no further questions but i have a a comment , er thinking about during our discussion and your defence (we're) starting with objectives for the research er in this area and then er carefully studying chapter eight i think there's a good link between the objectives and your research contributions it could (xx) and clear statements and er so from that point of view i think you have demonstrated that you clearly describe your research contributions in this area for hydraulic manipulators using water hydraulic er actuators and also (to do) in this area impedance control schemes , so there's a good link between the objectives and the final chapters and also some good er statements looking forward for research in the future thank you </S2>
<S1> okay </S1>
<S3> yes , okay i think we have come to the point where i would like to read what kind of statement we are giving to the institute of hydraulics and automation tampere university of technology , <READING ALOUD> the statement by the opponents concerning mr <NAME S1>'s thesis im- impedance control of water hydraulic manipulator for teleoperation applications </READING ALOUD> i'm not reading through whole this statement but taking parts of some some places er first few words about the background er <READING ALOUD> this thesis studies feasibility of water hydraulic rotary drive for position and torque control in implemented and impedance control scheme to handle the interaction between the manipulator and the environment and develop an efficient model based teleoperation method , and he focuses on control aspects of water hydraulic rather than mechanical design and material properties , the design the thesis gives an overview on the state of the art in water hydraulic tech- drive technology in impedance control and in human-machine interface for teleoperation , then he discusses on the differences and simirari- similarities between oil and water hydraulics and gives valuable knowledge on the manipulation ability of the water hydraulic manipulator compared to the more known oi- oil hydraulic one , theoretical studies are verified with experiments </READING ALOUD> and some words about er the aspects the contents <READING ALOUD> the contents of the thesis concerns three aspects the feasibility of water hydraulic rotary drive for position and torque control implementation of the impedance control scheme to handle the interaction between manipulator equipped with water hydraulic rotary drives and the environment and the development of an efficient model based teleoperation applications , as a conclusion the thesis indicates that the candidate has a deep knowledge on practical and pac- practical experience on the subject , the quality of the scientific contribution is very good and the author has done him- himself most of the work , also the theoretical analysis (al- we-) as well as the experimental study of the water hydraulic control problem are clear and easy to follow , as a whole the scientific contribution of this thesis exceeds the demand laid on the dissertation and also in the public defence of this thesis the respondent has coped very well with all the question placed on him to answer </READING ALOUD> and the statement <READING ALOUD> therefore we recommend the acceptance of this dissertation for doctoral thesis of for doctoral doctor of technology </READING ALOUD> </S3>
<S1> thank you i thank you for the criticism on the thesis , and if er anyone present still has some remarks or questions to ask concerning my dissertation i respectfully request that you ask for the <FOREIGN> kustos </FOREIGN> for the floor </S1>
