Dynomotion

Group: DynoMotion Message: 1740 From: bradodarb Date: 8/28/2011
Subject: Kinematics Configurations
Hello Tom,

First, sorry everybody invloved for the spam episode. My hotmail was hijacked for a couple hours but is back under control. I just wanted it to be known that it was an issue with hotmail and not my machine so rest assured the c# libraries are virus-free.


Ok, back on topic:

I was running some scenarios through my head this morning and thought of parallel kinematics.

My initial question would be:

Does the kinematics class have hard-realtime constraints?

Thank you,
Brad Murry
Group: DynoMotion Message: 1750 From: bradodarb Date: 8/29/2011
Subject: Re: Kinematics Configurations
The reason I am wondering if the configureation of the kinematics is required in realtime(vs set sand forget) is I would like to look at allowing kinematics models to be defined on the c# side and used on the Kflop side

-Brad Murry
--- In DynoMotion@yahoogroups.com, "bradodarb" <bradodarb@...> wrote:
>
> Hello Tom,
>
> First, sorry everybody invloved for the spam episode. My hotmail was hijacked for a couple hours but is back under control. I just wanted it to be known that it was an issue with hotmail and not my machine so rest assured the c# libraries are virus-free.
>
>
> Ok, back on topic:
>
> I was running some scenarios through my head this morning and thought of parallel kinematics.
>
> My initial question would be:
>
> Does the kinematics class have hard-realtime constraints?
>
> Thank you,
> Brad Murry
>
Group: DynoMotion Message: 1753 From: Tom Kerekes Date: 8/29/2011
Subject: Re: Kinematics Configurations
Hi Brad,
 
Not quite sure what you mean here.  Nothing on the PC is really real-time and I don't see how that relates to set and forget?  Not many Users seem to be using kinematics so I think we need a few more examples before we can decide how flexible it needs to be or what needs to be exported.  But basically it currently needs to be implemented a separate class that overrides some of the CKinematics members.  Mainly just two functions that map CAD space to Actuator Space and one to do the reverse.  Actually if one is provided it can compute the inverse numerically so only one may be needed in some cases.  Actuator to CAD is only really needed for Status and DROs for current position displays and plots.  But status is somewhat real-time.  Another relationship is that current position (robot pose) can influence accelerations and velocity constraints in CAD space which in turn influences Trajectory Planning.
 
Hope this addresses your issue.
TK

--- On Mon, 8/29/11, bradodarb <bradodarb@...> wrote:

From: bradodarb <bradodarb@...>
Subject: [DynoMotion] Re: Kinematics Configurations
To: DynoMotion@yahoogroups.com
Date: Monday, August 29, 2011, 6:16 PM

 

The reason I am wondering if the configureation of the kinematics is required in realtime(vs set sand forget) is I would like to look at allowing kinematics models to be defined on the c# side and used on the Kflop side

-Brad Murry
--- In DynoMotion@yahoogroups.com, "bradodarb" <bradodarb@...> wrote:
>
> Hello Tom,
>
> First, sorry everybody invloved for the spam episode. My hotmail was hijacked for a couple hours but is back under control. I just wanted it to be known that it was an issue with hotmail and not my machine so rest assured the c# libraries are virus-free.
>
>
> Ok, back on topic:
>
> I was running some scenarios through my head this morning and thought of parallel kinematics.
>
> My initial question would be:
>
> Does the kinematics class have hard-realtime constraints?
>
> Thank you,
> Brad Murry
>
Group: DynoMotion Message: 1755 From: bradodarb Date: 8/29/2011
Subject: Re: Kinematics Configurations
Hello Tom,

Sorry, I was not too clear there...
I understand with the possible exception of WinCE we will have no deterministic processing available and thus no hard real-time capabilities under Windows.

I guess what I really want to know is if the points are still buffered in the kflop does the overriden kinematics class get used from the device side or is it just for point generation?

If there are no real time requirements, I can implement a generic export template kinematics class and expose it to the c# side.

Thanks,

Brad Murry


--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Brad,
>  
> Not quite sure what you mean here.  Nothing on the PC is really real-time and I don't see how that relates to set and forget?  Not many Users seem to be using kinematics so I think we need a few more examples before we can decide how flexible it needs to be or what needs to be exported.  But basically it currently needs to be implemented a separate class that overrides some of the CKinematics members.  Mainly just two functions that map CAD space to Actuator Space and one to do the reverse.  Actually if one is provided it can compute the inverse numerically so only one may be needed in some cases.  Actuator to CAD is only really needed for Status and DROs for current position displays and plots.  But status is somewhat real-time.  Another relationship is that current position (robot pose) can influence accelerations and velocity constraints in CAD space which in turn influences Trajectory Planning.
>  
> Hope this addresses your issue.
> TK
>
> --- On Mon, 8/29/11, bradodarb <bradodarb@...> wrote:
>
>
> From: bradodarb <bradodarb@...>
> Subject: [DynoMotion] Re: Kinematics Configurations
> To: DynoMotion@yahoogroups.com
> Date: Monday, August 29, 2011, 6:16 PM
>
>
>  
>
>
>
>
> The reason I am wondering if the configureation of the kinematics is required in realtime(vs set sand forget) is I would like to look at allowing kinematics models to be defined on the c# side and used on the Kflop side
>
> -Brad Murry
> --- In DynoMotion@yahoogroups.com, "bradodarb" <bradodarb@> wrote:
> >
> > Hello Tom,
> >
> > First, sorry everybody invloved for the spam episode. My hotmail was hijacked for a couple hours but is back under control. I just wanted it to be known that it was an issue with hotmail and not my machine so rest assured the c# libraries are virus-free.
> >
> >
> > Ok, back on topic:
> >
> > I was running some scenarios through my head this morning and thought of parallel kinematics.
> >
> > My initial question would be:
> >
> > Does the kinematics class have hard-realtime constraints?
> >
> > Thank you,
> > Brad Murry
> >
>
Group: DynoMotion Message: 1757 From: Tom Kerekes Date: 8/29/2011
Subject: Re: Kinematics Configurations
Hi Brad,
 
Yes that should be possible.  The Kinematics converts the CAD Space segments to Actuator Space segments on the PC just before they are downloaded and buffered in KFLOP.  KFLOP then later performs the 8 axes real-time linear interpolation in Actuator (motor) space.
 
TK  

--- On Mon, 8/29/11, bradodarb <bradodarb@...> wrote:

From: bradodarb <bradodarb@...>
Subject: [DynoMotion] Re: Kinematics Configurations
To: DynoMotion@yahoogroups.com
Date: Monday, August 29, 2011, 7:34 PM

 
Hello Tom,

Sorry, I was not too clear there...
I understand with the possible exception of WinCE we will have no deterministic processing available and thus no hard real-time capabilities under Windows.

I guess what I really want to know is if the points are still buffered in the kflop does the overriden kinematics class get used from the device side or is it just for point generation?

If there are no real time requirements, I can implement a generic export template kinematics class and expose it to the c# side.

Thanks,

Brad Murry

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Brad,
>  
> Not quite sure what you mean here.  Nothing on the PC is really real-time and I don't see how that relates to set and forget?  Not many Users seem to be using kinematics so I think we need a few more examples before we can decide how flexible it needs to be or what needs to be exported.  But basically it currently needs to be implemented a separate class that overrides some of the CKinematics members.  Mainly just two functions that map CAD space to Actuator Space and one to do the reverse.  Actually if one is provided it can compute the inverse numerically so only one may be needed in some cases.  Actuator to CAD is only really needed for Status and DROs for current position displays and plots.  But status is somewhat real-time.  Another relationship is that current position (robot pose) can influence accelerations and velocity constraints in CAD space which in turn influences Trajectory Planning.
>  
> Hope this addresses your issue.
> TK
>
> --- On Mon, 8/29/11, bradodarb <bradodarb@...> wrote:
>
>
> From: bradodarb <bradodarb@...>
> Subject: [DynoMotion] Re: Kinematics Configurations
> To: DynoMotion@yahoogroups.com
> Date: Monday, August 29, 2011, 6:16 PM
>
>
>  
>
>
>
>
> The reason I am wondering if the configureation of the kinematics is required in realtime(vs set sand forget) is I would like to look at allowing kinematics models to be defined on the c# side and used on the Kflop side
>
> -Brad Murry
> --- In DynoMotion@yahoogroups.com, "bradodarb" <bradodarb@> wrote:
> >
> > Hello Tom,
> >
> > First, sorry everybody invloved for the spam episode. My hotmail was hijacked for a couple hours but is back under control. I just wanted it to be known that it was an issue with hotmail and not my machine so rest assured the c# libraries are virus-free.
> >
> >
> > Ok, back on topic:
> >
> > I was running some scenarios through my head this morning and thought of parallel kinematics.
> >
> > My initial question would be:
> >
> > Does the kinematics class have hard-realtime constraints?
> >
> > Thank you,
> > Brad Murry
> >
>