Detailed Design


Since the project was so open-ended, we decided that the feedback from our sponsor should be our determining factor. We held a conference with him, and discussed our observations with each model.
For the Flemons’ model, we thought it looked very interesting and was the best visual portrayal of tensegrity. It was also the original recommendation of our stakeholder. However, after constructing the prototype and experimenting with the cord tension for motion, due to the abstract shape, it wanted to roll over on the side being tightened. This was due to the decreased length at the spot and no support at the front or rear. If more links were used, this may have solved the issue, but that would become expensive and still wouldn’t be a sure solution. An alternative solution would be to add on an additional leg to the front and rear vertebra for stability; this idea was dismissed fairly quick, too.
For Danny’s model, it didn’t appear very tensegrity at first. Upon explaining the actuation of it, however, the sponsor was able to see how it used the principles of tensegrity. Our sponsor also pointed out that this model would be very collapsible. Since the idea for this project is a proof of concept, this would be very beneficial in validating space exploration capabilities.
The 3 spindle model was discussed as well, but the use of springs wouldn’t really fit with the tensegrity idea and this concept was with no surprise quickly dismissed.
Another structural issue we wanted to discuss was the placement of actuators. The project description specifies no base plate and the use of in-line actuation. We asked if that meant the actuators needed to be suspended on the line itself, or if the actuators could be attached to the vertebrae frame. He clarified attaching to the frame was okay.
Ultimately, he recommended we run with Danny’s model. He made up his mind fairly quickly and we felt it would be the best to specialize in, too. In addition to being the best liked approach, there was an added bonus: by placing a fourth actuator at the tip of the tetrahedron, it could tighten or loosen all three of the inner cords evenly which would allow an accordion/inchworm movement as well as help distribute the force as needed.


Upon choosing the form of the structure, some requirements were made to make sure it would work:
  • The frame would need to be easy to fix of replace a portion. If something breaks, we want to be able to fix it quickly without having to re-machine the whole frame.
  • The actuators need to be attached to the frame and be in-line with the cords. As all of the components for the actuators are still being developed and tested, the machining cannot be held up to long; thus, there needs to be space for actuators and an easy way to attach them.
  • All cords need to be the same length when at the same tension. Even using force actuation, the cords must be the same length structurally and have the same level of pretension. This would also allow easy future adjustability.
  • We cannot route the cords. That was an issue from last year that the sponsor would like to avoid this year. He wants to be able to see clearly what each wire does.

Current Base Ideas

To make attachments easy, it would be nice to have at least one flat surface for each vertebra. The most suitable place for this was the base, which is the connection of the legs opposite of the tip of the tetrahedron. The concept is shown below.

Figure 4: The base of a vertebra.

As shown in Figure 4 above, there should be plenty of room for three actuators. They will be placed on the back side of the base. This is to help add balance to each vertebra as well as to prevent the components from being in the way of actuation. Note, this will limit the range it can collapse inside itself, but currently the range of collapsing is limited by the size of the actuators in the nose cones, as they cannot be allowed to collide. So this would be a way to take advantage of that empty space. Still it must remain small for slithering type motion as that will not be restricted by the nose cone actuator size.
In addition, there would be room to accommodate for controllers and mini breadboards for circuitry if needed for future controlling. Basically, it should prove fairly versatile. The spindle holes could either be placed at the outer edge of the triangle tips or inner edge of the triangle tips.

Leg and Tip Attachments

There were several ideas of how best to attach the spindles and tip of the tetrahedron. We thought it best to make the tip its own separate part and the spindles separate as well, as this would allow easy replacement if needed.
For the tip, there were two main ideas. One, make it a simple ball shape; two, modify the ball shape to accommodate for an actuator to be placed in the center. The later concepts are illustrated below to better visualize.

Figure 5: The modified nose cone. This is the back view to see the actuator placement.

And as seen in the figure above, the modified shape would allow for an actuator to be slid through the center. Basically, the tip is shifted towards the base more to maintain the same angle for the spindles. In addition, it should have a reduced weight compared to the ball. However, the ball shape would be easy to manufacture. It could be clamped down and simply move the angle and position of the drill bit to drill the holes and tap as needed.
For the legs, we thought it easiest to simply use cylindrical rods. As far as attaching those to the base and tip there were again a few ideas. One, thread the holes; two, use a pin to hold the position; three, use a bolt. All three concepts are shown below.
(Include all three ideas)
Figure X: a.) Using threads, b.) using pins, and c.) using bolts.

As seen in Figure X a. above, threads would give stability and be easy to machine. But due to the geometry, it would be unnecessarily difficult to make threading both ends worth it. It would require (whatever the two-way threading part is called) and even then it would be difficult to get all three spindles to tighten at the same lengths.
Pins would work for either end, but may not give as much stability. To make the pins work, a hole could be drilled through the base at the correct angle to ensure consistency and side holes could then be drilled through for the pin to be inserted. The pin configuration would also be easy to machine. However, the use of pins may omit certain placements of the actuator since they might physically be in the way of each other, whereas threaded would not require any external space.
For the bolt shown in Figure X c. above, it would give great support and be able to mount to any flat or rounded surface; the end of the leg would just need to be flattened or rounded accordingly. However, this spindle design might be difficult to machine. It may also run into threading issues or space issues based on how the bolt is secured down.
Ultimately we thought it best to use the modified nose cone to enable the use of an actuator there. Since threading is easy to machine and stable, we also decided to thread the end of the leg that would insert into the nose cone. For the other end of the leg, we will angle a hole through the base frame as noted with the pin approach, but we are hoping that no pin will be needed. By inserting the bars through first, then threading them into the tip, the rigidity and angle of the bar should prevent any further movement between the base and legs.

Attaching Cords and Length Adjustability
This applies to the cords running from the base of one vertebra to base of the next, as well as the cords running from the base to tip. There were two main approaches to accomplish this. One, run the cord through a peg/dowel and then wedge it in a hole in the base; two, use the same idea as a guitar tuner. Both concepts are illustrated below.
(Include a picture of both cord attachment ideas)
Figure X: a.) Using a dowel setup, or b.) using a guitar tuner setup to attach the cord.
As seen in the above Figure X, the dowel setup in a., it would be efficient; however, it would be difficult to machine, and press-fitting would also be problematic to adjust the cord length. As for the guitar tuner concept, it would be easy to buy the part, then simple enough to drill and tap a hole for it in the base. The cord could be wrapped about the tuner, making a sure connection, and it would allow easy cord length adjustment.
We decided to run with the guitar tuner concept. It seemed straight forward.
 Actuator Attachment and Placement
To actually attach the actuator to the base we had two main ideas. One, a horseshoe clamp; or two, a clamp attachment similar to a scope mount. Both can be seen in the figure below.
(Include both actuator attachment ideas)
Figure X: a.) Using the “…”, and b.) using the clamp attachment to attach the actuator.
The horseshoe clamp seen above would probably be the simplest approach. We already have a flat surface to attach it to and now that we have our actuator dimensions, we could machine it really quickly. However, the added bonus of the scope-type mount is that it could accommodate any changes in actuator dimensions. Though it hopefully shouldn’t be necessary to change actuators, but would just allow adaptability.
As for the actual placement of the actuators, the figure below demonstrates where they will be located.

Figure 6: The placement of actuators on the vertebrae structure. The smaller actuators are shown in red, while the larger actuators in the nose cones are shown in blue.

The spindle of an actuator will be placed at one of the numbered positions shown above. For example, an actuator would be mounted on the back of the frame at position 4 and would have its cord attach to the front of location 1. The actuator at the back of location 1 would then run to the front of the vertebra behind it, and so on. This same concept is the same for all actuators 1-6. As for the actuators in the nose cone positions 7 and 8, actuator 7 would connect to locations 1, 2 and 3 with a three separate cords; while actuator 8 would similarly connect to locations 4, 5 and 6, again with three separate cords.
To actuate, actuator 7 would unwind while actuators 1, 2 and 3 would all tighten, thus moving nose cone 7 inside of the vertebra in front of it. Then by using one way friction, when actuator 7 tightens back up and actuators 1, 2 and 3 loosen, the vertebra it is in would be pushed away and forward from it. This would simulate an inchworm or accordion-like movement.
To achieve slithering or snake-like motion, actuators 7 and 8 could be held constant. Then by tightening actuators 2 and 5 and similarly loosening actuators 3 and 6 on the other side, the structure would want to turn to the side. By alternating this motion, and again with the use of one-way friction, the structure would slither forward. Its speed and magnitude will be driven by the original length of cord and power of the motor.


There is still room for additional equipment such as the use of load cells or optical encoders. But we feel that it would be good to get a good head start on machining and assembling the prototype so that we can test our actuators and other equipment. This tends to be the best way to test out our ideas and make sure we are on the right track, as well as more readily notice any drawbacks we may encounter.
So our next steps are:
  • Machine
  • Assemble
  • Attach actuators and equipment to begin testing


After discussing the possibilities with our sponsor, we immediately ruled out the linear actuator since its performance was much worse than the other two.  We continued to consider the air muscles for a while since our sponsor liked them so much, but he did think that they might be somewhat impractical.  In the end, we decided to not use the air muscles because an air compressor would be too big to make the robot practical, it would need to be transported to
NASA Ames, and the air muscles would not be feasible in space.  That left the rotational motors, which seemed to have some of the strongest advantages.  Also, our sponsor mentioned that we could mount them on the structure rather than hang them in line with the force.  The only condition was that they exert force in the direction of motion only.  This made them even more feasible.

We came up with two types of rotational motors that we could use: servo motors and DC motors.  The advantages of the servo motors were that they had readily available code and that there were a lot of models available to choose from.  The disadvantages were that while we needed a force based control system for the robot, the force based controls for the servo that we tested were very inaccurate and it would not be very feasible to use the position controls to control output force.  These disadvantages led us to use the DC motor instead.

The DC motors advantage was that the input current is roughly proportional to the torque output.  This means that we could more directly control the torque using a current controller.  In order to implement this we will need to build and test our own controllers.  This is where the actuation of the tensegrity robot now stands.

Actuation selection was modeled in appendix A.


Our current design of the actuator system is to use DC motors that are directly connected to the frame.  There will be four motors mounted on each section of the structure with three of the motors holding the structure in tension with the other motor.  The motors will use spindles to wind and unwind the strings.  Motion will be achieved through changing the force that each of the two groups exert on each other in a synchronized manner.  Each of the motors will be controlled by a single current controller and it is hoped that a single microcontroller will be able to control several of those.  These will need to be mounted to the frame.  The power source will need to be connected to the robot by a tether.


Now that the electric motors have been purchased, future actuator work will be in refining the method of forced based control. At the half way point in the semester there are two different ideas for controlling the force that each motor applies.
These two types of force control do not alter the structure of the tensegrity and can be investigated as parallel sub-projects without inhibiting progress on the rest of the robot.


With the parameters of what we are supposed to have for our design, we are thinking that both load cells and the current controller are going to have to be incorporated. The parameters of the project says that the robot has to be forced based and the force has to be displayed at all times. This force based is going to come from the current controller and the displayed force is going to come from the load cells that would be attached to all the tendons in the design. With these two control methods we are covering all parameters.


Figure 8: Shows the current controller circuit board and how it is designed.

Figure 8 shows the way the current controller is set up. This design will be able to control the torque based on the fact that current and torque are proportional. We will control the current by calculating the current over a known resistor and voltage. Then a map of the current to the torque output of the robot will be made.


The first few weeks of the second semester of this project will be used to finish up the testing of both types of control. At the end of the first semester the best option appears to be the combination of both types of data acquisition. For the force control a current controller is the most efficient method. However, if load cells are also included then the tension can be accurately measured for every line rather than the torque applied at each motor.
Both circuits will be built by the team then ordered from a company that will give more compact and reliable circuits. Extra circuits will be purchased to insure that the robot can run for many years, as well as clearly labeled circuit diagrams that convey how each one is built.


We have decided that the most effective and easily incorporated one way friction would be a cross country or telemarked climbing ski skin. If it becomes necessary for us to have one way friction we will incorporate these climbing skins.


We have not modeled anything yet, but we feel the ski skins will be the best design for our project. Once we have our structure machined, we will attach the ski skins to the bottom of the base frame. We can adjust the foot print of the base frame to find the correct amount of friction from the ski skins.


If it becomes apparent that we will need one way friction in our robot, we will continue with one way friction design. The first step in our one way friction design will be to set up an experiment to determine both the static and kinetic friction coefficients in both directions of the ski skin. This will be accomplished with a couple of very simple experiments. The first experiment will be to find the static friction coefficient this will be done with a tilt table and a sample of our ski skin under a weight. The coefficient of static friction will be found by the tangent of the angle at which the block starts to slide. The coefficient of kinetic friction will be found by using an Atwoods machine. An Atwoods machine consists of a block or a sled with the surface on which you are testing the friction of; a frictionless pulley, a level table, and a hanging weight. For the Atwoods machine to work, photo gates will be setup a set distance apart to calculate the velocity of the block on its path past the photo gates. The full set up and experiment details can be found in “Experiments in laboratory Physics” by Yanko Kranov Experiment 5. Once the coefficients of friction are found, we can input them into our math model so we can map the position of our robot to the force that we apply in each of our strings.