When we think of human control systems we most often think brain. It may surprise some to learn that as much processing occurs in other body parts as what comes from the mind. Most are aware that sensory reactions such as retreating from a burn are triggered in the spinal cord. Musicians and others who depend heavily on kinetic motions are well aware of muscle or kinetic memory where actions are automatic and depend on the interrelation of tissues and nerves. Robotics would do well to incorporate this distributed processing to simplify CPU responsibilities.

I have to agree with you on this concept. I noticed years ago that at Stanford University as they kept developing their robot that they openly stated that the more intelligence they put into it the slower it responded. They had it negotiating hallways and doing all kinds of physical things. This brought to mind an artical that I had once read in a magazine.
Now whether it was true or not it made me think.
The author stated that some dinosaurs were so large that they had two brains, one in their heads and another that helped controlled they musclar movements.
The first thing that popped into my head is why don't we apply this dinosaur principle as I called it into our robotic designs. We could use one to control the arms, hands, etc with all of its special programming isolated from the rest. Another could be used to control its legs, feet, etc and all aspects of walking, climbing. Then the 'last' one (there could be more) could maintian order and control all of the others. The brain could give the signal to the legs to walk in a certain direction without having to worry about how the legs would execute this directive. Simplified system could lead to a much more advance system that what exits now.
There is tons of programming on vision, hearing, speaking, walking, talking, etc and if one PC tried to intergrate all into one massive program the robot might work but not very effeciently. However if each process was controlled by its own processor it would come closer to acting instaneously just like a human would.
So to me the 'dinosaur principle' is going to be a real key to the advancement of robotics beyond the 'toys' that now exist.
Bill
P.S.
I am both old and opinionated.

While we don't often think of it, the brain is not a single processor. We have defined areas of the brain that process various functions.

Computers are beginning to use multiple processors, but the divorce is not broad enough. These processors need their own memory, I/O, and control busses. Yet they need to have a common center of communication. The brain has the frontal and prefontal cortex.

If you study SAM you will find that it is capable of hundreds of separate processors that multitask, communicate to ever higher levels, and have a single supervisor at the top of the stack. I believe this closely corresponds to the structure, and even the methodology of the brain.

I remember reading years ago someone quoting that dinosaurs were so big that it was believed that they had two brains one of which helped to control their body movements. It might not have been true but it sent me thinking. Standford University has had a long histroy of working on robotics and they stated that the more that they programed the robot to do the slower it would run.
This made me think why not have seperate PC's each one preprogrammed to handle specific tasks such was walking, grabbing objects, etc. Then the main PC would interact with the others thus being able to do more than one thing at a time and speeding up the whole process.
I don't know about SAM or even where I can read about it but if it proposes anything like what I have just written I would be interested in reading about it.
Bill