A Dimarts 20 Març 2007 15:50, Jan Kiszka va escriure:
> Leopold Palomo-Avellaneda wrote:
> > A Dimarts 20 Març 2007 15:00, Jan Kiszka va escriure:
> >> Leopold Palomo-Avellaneda wrote:
> >>> ...
> >>> Ok,
> >>>
> >>> thanks for the answer. I have to addmitt that to me all the
> >>> documentation is a bit confuse for someone that begins in this area.
> >>> Although there are a lot documents, I'm a bit lost.
> >>
> >> That's what we are gradually trying to improve via the wiki. Work in
> >> progress, contributions of any form are welcome.
> >
> > a Readme first ...
>
> What information are you missing in the existing top-level readme?
Well first of all please I have to agree all you effort (and the others
developers) in the rtnet. To me the main problem is the begin. I'm not an
expert in realtime and I'm doing this for a work in my phD using only the rt
as a tool, not as target.
Mainly, although I have a rtnet configured and working I'm not sure what I'm
doing. Probably is my problem because the lack of knowledge in this area and
the nomenclature or the concepts are a bit unclear to me. Basically I need a
step 0 or step -2 to begin. In the top-level readme I found:
What is it?
....
... RTmac layer
I don't know what is a RTmac layer
...
and I don't know what is a TDMA
What are the requirements?
ok
How to install it?
ok
How to test it?
3. Check <PREFIX>/etc/rtnet.conf and adapt at least the following parameters:
RT_DRIVER, IPADDR, TDMA_MODE, TDMA_SLAVES.
maybe an example of configuration (or I didn't found it) with two boxes with
cross cable, tree boxes an a hub, n-boxes and a hup...
Special notes
3. Configure real-time NICs: rtifconfig rtethX up <IP>,
rtifconfig rtlo up 127.0.0.1
4. Add host routes to target nodes, either explicitly:
rtroute add <target IP> <target hw-address> dev <local NIC>
Or trigger an ARP handshake:
rtroute solicit <target IP> dev <local NIC>
in a normal net I don't need to add a node if it has a local address, maybe a
point that it's necessary in the rtmodule (or I have understand it)
[.....]
> >> If you picked the default setup via rtnet.conf, TDMA was activated at a
> >> cycle period of 5 ms, one transmission slot per node and cycle. Thus you
> >> get a latency of up to 2 x 5 ms.
> >>
> >> You can improve this by reducing the period or adding more transmission
> >> slots per cycle. If you only want to use the RTnet link for RT traffic
> >> and you have a collision-free media (cross-link or switched Ethernet),
> >> you could also run RTnet without RTmac/TDMA. Writing a specialised RTmac
> >> discipline (as a replacement for TDMA) is yet another option, but surely
> >> a more complex one.
> >
> > Ok, I only want a simple thing. The problem is that I need the fastest
> > low latency response as I can. I have only to send some bytes, no a lot
> > of info.
>
> Then you probably want plain RTnet without RTmac/TDMA.
so I throw away the tdma.conf file and I only wotk with the rtnet.conf.
> > You comment something about to modify the tdma.conf file. I have this
> > parameters:
> >
> > master:
> > ip 192.168.1.3
> > cycle 5000
> > slot 0 0
> > slot 1 100
> >
> > #slave:
>
> ^^^
> This comments out the slave section mark, thus breaking the syntax.
yes, I did a mistake!!!!!!
> > ip 192.168.1.3
> > #mac 00:C0:3A:25:01:OB
> > mac 00:A0:C9:1D:52:B9
> > slot 0 2400
> > slot 1 2200 2/2
> >
> > but I don't know if this are correct or not.
>
> The fragment above is for demonstration of the tdma.conf language. See
> Documentation/README.rtmac for details about what those lines express.
>
> >> However, it all melts down to scheduling your network traffic for hard
> >> real-time use, not to make it simply as fast as possible, but to make it
> >> fully predictable.
> >
> > Yes I know, but is a simple system: a PC with rtai and rtnet and a robot
> > controller with vxworks.
>
> This doesn't tell anything about the communication structure, the timing
> of your system, or the closure of the control loop (if there is any).
Well, I have a rtai box with orocos [1]. The idea is that here we will
implement a software controller of a robot. In the other part (the vxworks)
we have the controller of the robot. However we have a library that let us
send the torque to each joint and read the position of the jooint. The
controller says that has a close loop of 4 ms. So the idea is that in this
time, we could read the positions of the robot, send then to the
rtai-controller and read the torques to apply to the robot. All of this is
thinking that we can do the control in one milisecon in the rtai-controller.
Regards,
Leo
[1] www.orocos.org
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
RTnet-users mailing list
RTnet-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rtnet-users
