Much progress has been made in the last ten years in the modeling of animal locomotion. Better quadrupeds (like the famous BigDog series), bipedal robots, and walkers of many kinds have made headlines, but the snake has received little attention. Most snake-like robots are not free-moving, but more like tentacles anchored to a larger robot frame.
The five types of snake locomotion have been well documented, and artificial muscles of various kinds have improved greatly over the years. I'm using highly twisted bundles of polyethylene and nylon fibers, which are extremely strong, and contract and expand greatly (almost 5% of length) as they are heated/cooled. Actuating the muscles via electroresistive heaters within each bundle, and cooling via a fluid jacket allow relatively fast actuation (~10/minute depending on ambient temperature).
I believe true snake robots will allow for very new applications as a cheap true 'go anywhere' platform that can go into tiny spaces
Today tested the twisted synthetic fiber component of the artificial muscles. I was afraid enough heat to cause action in the muscle would be near melting, but it surprised me. I twisted eight feet of poly propylene filament under tension until it kinked along it's whole length in a pattern called inverted twist entry.
Liberal application of a heat gun at various points caused the fiber to raise the test weight (about 32 grams) up by 2cm per 15 cm of fiber heated, an impressive contraction AND lift capacity for less than a tenth of a gram of fiber.
More impressive, when allowed to cool, the fiber expanded and allowed all future contractions to be just as energetic.
According to a paper written on using synthetic fibers for these uses, nylon can be directly actuated via high voltage electricity, which will be my next test material. Bundles of oppositely twisted fibers should be stable, and based on these experiments cool quickly in open air, so dense heat wicking material or fluid jackets should provide enough heat transfer to allow relatively fast actuation.
The controller for this is probably going to be a teensy driving a 48 channel pwm controller running off a battery chain down the middle of the robot.