2000 oz at the shoulder,and springs or tendons does help fairly easy to build 2000 oz servo ,at the shoulder it takes the most load, like it says at society of robots article (witch is very good) other arms you need is about 1/3 less on servo torque

can buy a 2000 oz servo but at $800 to $1000 each

will be posting my hand claw design of my omnibot as soon as the parts come in (2 days) and then build it holding a beer i also bought a plastic hand to look like a real hand to add latter,but it uses plastics parts ,so looking at machining them out of aluminum ,fingers are the hardest to make and control

Look at these
rubber bans & compressed air

Thomas this one uses tendons and motors...

tendons & motors

Different type of arm but interesting.
how robot arm works

Rgordon, To answer your questions from 6/27... My arms are only going to have two flexible joints in each. The part that I think of as a shoulder will only slide in and out, extending just past the arm socket. There will be a wrist beyond that. I figure that when Bob May wore the original costume; he would have not gotten his shoulders out into the arms anyway. The two joints should be able to more than replicate any moves that the original robot had. I made a rough CG model. I do not have a cad program to put in all of the details, but this should give you an idea. Where the red and blue cables meet the motors, there will actually be a section of bicycle chain that fits over a sprocket on the shaft. Of course, by the time they get to the pulleys, it will be plastic coated cable. The green piece will be a universal joint. Unlike the picture, it will be attached to both sections. I already have a few small ones, but the ones that fit into a ratchet and socket set would work. As the red and blue cables move, they will pull the forearm in any direction. I intend to put the sliding potentiometers on the flat board under the cables and attaching them somehow to keep track of their positions. The cable to bend the wrist will be similar to a brake cable for a bicycle, but heavier duty. I didn't show the other end, but it will be a simple lever on the end of the motor. My picture shows the whole inside to look like wood. I did that just for clarity. The flat board will probably be MDF. The shaft to the universal joint will likely be aluminum. (It does not necessarily have to be round either). The forearm will be PVC. The wrist will have to be fabricated. I will need a motor to rotate the wrists. I may use actuators to open and close the claws, or I might use bicycle cables there as well. In my case, and probably with Magnus too, the whole mechanism will have to slide in and out. I will attach it to drawer slides between the arm sockets and the back of the torso. Just picture the flat piece moving right out to the arm socket with everything else sticking out into the rubber arm bellows. To answer your last question, I do not think that I will have a central support system. The arms will be built right into the torso. The electronics will be built into a unit that can slide in and out of the top.

I am sorry for such a long post. Perhaps there are ideas here that others can use. I also have to add that I have not tried any of this yet. It may be a miserable failure. In a few weeks, I have some time off and I will be trying it out.

Incidentally, sfoy, I found the links that you have been sharing fascinating. The goal of my bumbling B9 is just to animate him in some crude fashion. I do love to study real robotics however. Perhaps some day I will be able to work on something more sophisticated.

i guess the motors you are using are car window motors,they do draw a lot and lot of current i bought for them ,2 to have on hand for spares or other robot design and 2 for for my hand design for a very big robot,waiting to get a lathe to build the arm and hand with fingers fngers with be the hardest to control,in my design i have 3 itx computers ,one for navigation ,one for sensors and servo,s other for the main brain and software,body and wheel base done as head design did some tests at work and from the tests using a torque tester and weight looks to handle and lift 25 lb or more

robotmaker The motors are actually windshield wiper motors. I've never scientifically tested the torque on them, but they have a lot of guts. If you ever had the wipers come on while trying to clean the snow off the windshield you'll know what I mean. My arms will be geared down much slower than wipers move as well. They do draw a lot of power. Because of the size of my robot, I am putting in two deep cycle marine batteries in him. I have 900 amps of cold cranking power. That ought to keep him going for a while. Besides, he probably won't be moving his arms much. Maybe an occasional handshake, or wave. I do not foresee him taking up tennis, or anything like that.

wow one very very heavy robot,marine batteries are the worst for robots,mostly acid and corrosion seen many of them use in robots and go bad alot gel cells are rated #1 for robot designs, i guess you are using wheel chair motors for the base thats what i am using in my big robot design,work out the current to be very low with all sensors,boards plus window power lift motors on arns and other current servo's and design is so well lasts over 4 hours on 35 amp gel battery that does weigh to much,and still handle a over extra 130 lb load

and windowshield wiper motors dont have alot of torque as you think,but do have more then most servo's i think 700 oz i saw somewhere online they only moving a wiper blade back and forth,not like a window power motor that has to lift a heavy glass window

wiper motors don't just go back and forth. They go around the same as a window motor.