Structural concepts
Tom Flemonsí structure was the initial recommendation for the structure of this project. As is seen, all of the vertebrae members are isolated from each other due to tension in the strings. So by reducing tension in certain strings and increasing it in others, you can get the vertebra to rotate about each other slightly, giving the slithering motion discussed earlier. Alternatively, by increasing tension in both parallel strings, you can get the vertebra to move closer to each other similar to an accordion or an inchworm.
To understand how to coordinate these strings, we had to create a prototype of this model. The other benefit of prototyping this was being able to discover possible difficulties that may have otherwise been unnoticed. Below can be seen our prototype of Flemonsí model.

Figure 2: Our prototype of the Tom Flemons model.
As seen in Figure 2, holes were drilled into the wooden balls equidistant from each other. Wooden spindles were then glued into each of the holes. An elastic cord was then glued from each spindle to its adjacent neighbors. In this model, the actuators would be placed on the outermost cords as illustrated by the red arrow, meaning a total of eight actuators for the 3-vertebra structure. By not allowing the other cords to change tension, it would create an axis as denoted by the green dashed line. If the outermost cords are tightened and loosened, the vertebra would then rotate about this axis in a snakelike slithering motion. For more universal movement, actuators could be placed on each cord, but for the scope of this project, it would become extremely expensive and extremely complicated to control.

Danny's model
This concept was coincidentally named after the member of our group that initiated the prototype. It is ultimately the same tetrahedral shape as produced by Flemons, but there is no center; rather the spindles form the tetrahedral shape and are connected together at their ends. The initial prototype was constructed out of toothpicks, erasers and rubber bands. From that initial concept, the final conceptual prototype can be seen below.

Figure 3: The final conceptual prototype of Dannyís model. The original prototype of this model can be seen at the top of this picture with the bright colored erasers.

As can be seen from Figure 3, the legs form the frame of each vertebra. The three innermost cords as shown by the red arrow are what suspend the tip of the tetrahedron. The outermost cords denoted by the green arrow stabilize and suspend the other ends of the tetrahedron. When held in any direction, the structure maintains its shape. Similar to the Flemonsí model, actuators would be placed on the outermost cords and by tightening and loosening the outermost cords the vertebra can pivot about the tip of the tetrahedron. This model isnít quite as constrained as the Flemonsí model, since there is no particular axis formed for rotation; however the tradeoff is that it would require the coordination of the three actuators rather than the two as in the Flemonsí model.

 3 Leg Model
This would be similar to Flemonsí model as well, except with only three spindles, all of which would still be equidistant from each other on the same plane. This concept seemed easier to actuate and manufacture. The hope was that there would be some way to attach the strings that would keep it stable with tension; but upon fabrication of the spindle setup, we couldnít find any way to keep the vertebra apart from each other without some sort of spring


In the early part of the semester we brainstormed three different ways to animate the tensegrity structure.  They were: air muscles, linear actuators, and rotational motors.

Air Muscles
The advantages of air muscles were that they were similar to biological systems, could exert a sizable force, and could change their length very quickly.  We also had a large amount of air muscle equipment available and faculty that was willing to help us.  The disadvantages were that its change in length was small compared to its overall length, it would require a bulky and expensive air compressor, and that such a system would not work well in outer space.

Figure X: Air Muscles. The upper is compressed with air, the lower is at rest after air is released out.

Linear Actuators
The advantages of the linear actuator were that it produced force in line with the desired motion and that it did so without the use of compressed air.  The disadvantages were that it did not change its length much compared to its static length and it was not able to produce much force.

Figure X: Internal view of a linear actuator.

Rotational Motors
When considering using rotational motors as a means of actuation, we came up with two possibilities: DC motors and servo motors.  The benefit of the DC motor was that the supplied current was directly proportional to the output torque, while the benefit of the servo motor was that there is a lot of software support for control.  We decided to experiment with both types of motor to determine which would be better for the project.  We obtained a servo and a controller and began testing to see if we could directly control the torque.  The torque control that we could obtain was very imprecise and hard to use.  We decided to forgo use of the servo motor in favor of the DC motor.

Mounting of the rotational motors has undergone some design changes.  At the beginning, we were under the impression that the motor needed to be suspended between the points where the force is applied.  One of the ways to achieve this is shown in the following figure below. After talking with our sponsor, we found that we would instead be able to attach the motor directly to the frame.

Figure 7: Rotational motor with an attached gear box on top to rotate the force direction.

The advantages of the rotational motor were that it would be easy to obtain, would produce a large change in length, could exert large forces, and would not require compressed air.  The disadvantages were that we would need to convert the rotational force into a linear force, and that it would be slower to respond than the air muscles.

The first method is simply proportional feedback control using a load cell to have the motor converge upon a desired force. This method will allow the tension in each line to be known at all times because the load cells will be collecting data even when the robot is not moving.
The second method of control that is under consideration is a current controller. The current controller will control the amount of torque that each motor applies rather than the tension in each line. However, a force will be the input into the user interface then computer will convert that force to a motor torque. With the current controller a circuit measures how much current is being applied to each motor, and when the desired the desired current is reached then the circuit will stop at that current causing the motor to apply the desired amount of torque.

We have foreseen the possibility of needing a one way friction source to make our robot move. This was realized in our research of the movement of snakes. We have discussed multiple ways of achieving a one way friction. Some of the ways that were discussed were a ramp style one way friction, a homemade linear bearing as shown in Figure 8, a tooth brush angled with all the bristles cut off in one direction, and cross country or telemarked climbing ski skins.
Figure 9: Linear one way bearing.
In Figure 9, the linear one way bearing would be pushed forward easily but cannot be pushed backwards because of the angled brushes. They would cause friction between the structure and the surface it would be walking on. This would ensure that the robot moves forward instead of backwards.
Figure 10: Side view of a rubber ramp for a one way friction concept.

The rubber ramp would move forward but not backwards just like the linear bearing. This concept might be a little easier to manufacture. It causes friction when it is trying to be pushed backwards because of the angled pieces of the design.