Robot arm, running control cables through joints -how to.

Here's the problem;
Say you have a robot arm which has several joints along it's length, each one is driven by a pair of steel cables. The other ends of the cables are connected to gearmotors which are located off the arm, to keep the weight down. So the cables which drive the farmost joints need to run through the preceding moveable joints, but the path length of each cable must not change. ie, if the cable were to stretch or become slack 'round the bends' it would affect the positioning of joints further along the arm.
It can be done using bowden-type cables of course, with an inner cable and an outer sheath, but this is not a very elegant solution. You can also try and route the cables through the centre of each joint, but unless they will bend to a very tight radius, the length still changes.
Is there a simple geometric solution to this, whereby cables can be routed around a joint, and as it flexes the path length remains constant? The Barrett Technology WAM arm appears to work this way, but I can't see how. I have thought up one solution using a pair of counter-rotating rollers, but I don't think the WAM works this way, and I'm sure there's a better way.
Any descriptions or links, especially with diagrams, much appreciated.
Andy.

    --Go to carlpisaturo.com and see how he's done it. Saw a drawing he'd made in the class I took; the solution is non-obvious; i.e. there *is* a trick to it..

steamer wrote:
  Quoted Text Here  
Thanks for the pointer on this site, it is EXTREMELY cool!
Bob

  Quoted Text Here  
Thanks, I'd actually already seen this site. His solution is to run the cables straight through the centre of the joint and bend them round as small a radius as possible when the joint moves. He admits this is not a perfect solution.
What I'm looking for is a mechanism that moves with the joint, compensating for the path-length difference.
Andy.

Andy wrote:
  Quoted Text Here  
Can't you just do it programmatically? What's an importance of hardware solution?
An example would be human fingers.

A pull/pull system with the actuator end floating might do the trick
David - who knows bugger all but likes to think about stuff
  Quoted Text Here  

Yeh, like steel cabling but lubricated with conductive grease
----------------------------------------------------------------------- Ashley Clarke -------------------------------------------------------
message
  Quoted Text Here  

Disclaimer: I don't have much hands-on experience with these methods. However, I have seen other systems using them.
Sprockets and chains can be used if you don't need many actuators. In the trivial case, a single chain connects the motor to a sprocket on a parallel joint axis; the chain length does not change as the motor and joint rotate. In the next case, one chain connects the motor to a free-wheeling sprocket on the first joint; another sprocket is fixed to this one, and a second chain connects it to a sprocket on the second joint. This easily iterates to cover N joints, but the number of sprockets and chains quickly becomes excessive.
If the joints don't need to rotate more than 180 degrees, a cleaner method can be implemented. A (vinyl?) coated cable is used with pulleys instead of using sprockets and chains. As in a chain, the cable forms a closed loop. One end of this loop is firmly attached to the motor. The loop then criss-crosses between free-wheeling pullies collinear with the joint axes. Finally, the other end is firmly attached to a pulley on the joint to be actuated. The 180 degree restriction comes from the need to attach the cable to pulleys on both ends; it could be alleviated by attaching the cable to a chain/sprocket system on both ends. Chains aren't used the whole distance since they do not handle sideways displacement well.
In either method, the final joint angles move as intermediate joints change. However, this is often desirable because the joint motion preserves the absolute joint angle (e.g. a link stays horizontal as the shoulder moves up and down).
Yet another approach is to use rotating shafts to transmit motion. At each joint, these shafts terminate with a flexible joint (universal joint or the like). This joint then connects to another shaft which transmits power to the next joint, and so on. The flex-shaft which comes with a Dremel tool can also be used for this purpose. Since its intended for light torques, heavy gearing at the output might be required. This class of approach generally cannot handle sharp joint angles.
Later, Daniel

?.?@?.com wrote:
  Quoted Text Here  
How about dual pullies?
Basically any cable only goes the length of a single joint where it is wrapped around one of a pair of linked pullies (or sprockets). These pullies would go through the center axle of the joint.
I don't think that this would be the most efficient way of doing things, but you wouldn't have to worry about routing. -- D. Jay Newman

Hi Andy,
You can take a look at my website aivision.com and see how running spectra (fishing or kite string) lines through copper platted spring coil can animate a robotic hand with 18 degrees of freedom.
Brad Smallridge aivision.com

?.?@?.com wrote:
  Quoted Text Here  
Read "Robot Evolution" by Marc Roscheim. His big picture book of robot arm internals and designs is just what you need. Roscheim makes a big point of getting all the cables inside the arm, an important feature in industrial robots.
http://www.wiley.com/WileyCDA/WileyTitle/productCd-0471026220.html