[Playerstage-developers] Arm interface(s)

[Daniel A. <bi...@kt...>]

 > Date: Thu, 31 Mar 2005 10:13:48 +0200 (CEST)
 > From: Herman Bruyninckx <Her...@me...>
 >
 > Maybe this is a good opportunity to really start the discussion/work on the
 > integration of Orocos with Player :-) Orocos (<http://www.orocos.org>) has
 > already the code for force controlled manipulators (_and_ working
 > applications!), and the integration into the mobile platform infrastructure
 > of Player would be a great competitive advantage of open source :-)
 >
 > Discussion of interfaces is the right place to start.

 > Date: Thu, 31 Mar 2005 10:20:12 +0200
 > From: "Kratochvil, Bradley Eugene" <kra...@ir...>
 >
 > Have you thought about using the Position3d interface for this arm?  The =
 > reason I ask, is that we currently have a SCARA style arm that we've =
 > almost finished the driver for using the Position3d interface, and I =
 > think it's going to work quite well.  We also have a bit of a special =
 > arm that's been working on the same interface for quite som time now.  I =
 > think defining an interface for robot arms may be a bit tricky.  If you =
 > do end up working on a new interface, it would probably be good to base =
 > it heavily on the position3d one and go from there.

I'll address these two issues at once. In my oppinion a complicated device such 
as a robotic arm can not be represented (in a natural way) by a single 
interface. The position3d interface would certainly be a candidate when it 
comes to controlling the arm's end effector using the arm's inverse kinematics. 
This is not the only use of manipulators, for instance we might want to support 
joint control where each DOF of the arm is controlled directly. This might seem 
like a problem since different arms have different kinematics and/or a 
different number of DOFs and thus the device abstraction would be lost. This 
does not need to be the case if the joint configuration is computed by a 
path-planning algorithm that has knowledge of the arm type. Thus an interface 
for supporing movement in joint-space would be required. Extending this there 
could be need for interfaces supporting force control, which I assume is what 
Herman is hinting at.

Clearly there is no obvious way to solve all of this in a simple way, therefore 
I believe a discussion is in place. So to summarize:
i) I believe the position3d interface is suitable for controlling a mainpulator 
in the workspace.
ii) I believe that at-least a interface for supporting control in joint-space 
is required.
iii) I believe that additional interfaces supporting e.g. force control could 
be usefull.

/Daniel