Sophmores enrolled in PoE are expected to "work in small multidisciplinary teams to design and to build a mechatronic system of their own choosing," reads the description on the PoE website.

PoE provides roughly $300 to each team for material costs and expects students to work 8 hours a week outside of class. However, the nature of our project required additional funding and time, so as a group, we added more credit-hours in the form of independent research at Olin. The Olin Self Study, Independent Study, and Undergraduate Research Board (OSSISURB) provides more funding on a per-group basis, the amount of which subject to need and of course, budget. From OSSISURB and Professor Pratt we received an additional $2000, bringing our total funding to $2300. (This figure includes purchasing of raw materials for next semester.)

Coming into the project phase of PoE, our group had an advantage over our classmates since we began work towards the beginning of the semester, doing preliminary research and component trade studies in addition to the assigned classwork. By mid-semester, when the class had completed the in-class introductory labs, we had completed our trade studies and had our Phase One snake fabricated and undergoing testing.

During our initial research we came across a multitude of modular snake designs, where each segment of the snake would be jointed to the next at a fixed location. Rather than repeat the direct-drive solution, we came up with a few unique ideas of our own, and eventually settled on the linkage design presented on this webpage. The design emphasizes the continuous nature of a snake by using an actual spine, in this case, flexible Lexan rod, and using servos and linkage arms to control the curvature of the spine.

By controlling curvature instead of joint angle, we feel we have approximated the actual mechanism at work in a live snake, where muscles attached to the ribs control curvature of the spine. By the end of the fall 2005 semester, we had created 4 generations of planar (2D) snake designs. Our final deliverable for our Principles of Engineering class was a 4 foot long, tethered (for both power and control), 4th-generation planar snake (Phase Four).

Chris and Gui went on to build a three-dimensionally actuated sidewinding snake with some help from the rest of the team. The next potential step would be to investigate three-dimensional, completely untethered robotic snakes.

All of the control algorithms are run in C on a laptop computer which serves as the USB host. Commands are sent to a PIC microcontroller, which are then transmitted over I2C to the servo-controller.

The HS-81MG servo from Hitech serves as a transducer between electrical and mechanical systems, translating control logic to physical spine curvature. Please see the individual snake phase descriptions for a more detailed explanation of how we used the servos.

This particular PCB was designed to replace our breadboard. It essentially has a 26-pin DIPP socket for a PIC18F2455 which translates USB traffic to I2C commands to the SD-21 servo controller. It also has a number of USB-controlled (through the PIC) LED status indicators for diagnostics and debugging.

We used ExpressPCB Design to manufacture the PCB.

One next potential step would be to get a gumstix running on the snake itself along with batteries for power, allowing us to control the snake without a tether over 802.11b WiFi.

The gumstix embedded computing platform is essentially a Intel XScale procesor running a linux kernel. Our model, the connectix 400, has a 400MHz XScale processor, Bluetooth connectivity, as well as expansion cards for servo control, general I/O, both digital and analog, and 802.11b connectivity.

The Phase One and Phase Two snake use a very simplistic serpentine gait algorithm. The simplest form of serpentine motion in a snake a that of a sine wave. Since servos are position-controlled, we update each servo on the snake with a sine wave, changing in time with a phase shift between each servo. The net result is a ripple effect in the snake, much like what you see in a Slinky.

The control algorithm for snake phases Three and Four is best viewed as a type of brute-force algorithm for the computation of joint angles. In software, a path is laid out on the ground in front of the snake, and for each head position along that path, the software computes the joint angles needed to keep each node on the path. In this way, the snake can follow the path on the ground (in "world-space") while each node only moves tangential to the path.

All of the parts were cut from sheets of 1/8" Delrin by a laser cutter. All of the servos were mounted on the each plate so that their output shafts were exposed on the underside of each plate.

During this phase, we managed to connect all of the components together, getting the PIC mediating between the servo controller and the host computer over USB and I2C.

A sinuosoidal position signal was sent to individual servos, with a phase shift between each servo. This signal resulted in a undulating serpentine motion in the snake, but wiring and mechanical problems made it unlikely that it would have actually movied.

Initially, all ribs were assembled using the thermal press; however, after spending 30 minutes per rib for 6 ribs, we decided to use 440 screws in tapped holes to fasten the ribs together. In the end, this cut our assembly time down by a factor of 10. In an effort to conserve raw material, the final revision for this phase used both types of rib.

The servo-bearing ribs (driving ribs) were heat-staked and the non-driven ribs (dead ribs) were fabricated using screws. This prevented us from having to fully re-cut the whole of the snake from Delrin, yet still minimized assembly time.

Both electrically and computationally, the Phase Two snake is essentially identical to the Phase One snake.

There are two independent wheels and axles on each live rib, which permits individual ribs to turn. Previous snakes attempted to combine two wheels on one axle, which forced wheels to slip as the snake attempted serpentine motion. The ribs were designed with ease of assembly in mind. Snap-together plastic assembly techniques were investigated and successfully implemented, and as a result each rib assembled in less than 2 minutes. With all of these new developments, we could install and remove servos without disassembling the entire snake, which was previously impossible.

The electrical system on the Phase Three snake worked very well. Abandoning 22 AWG solid core for 26 AWG stranded core eliminated the flexibility problem. Additionally, the wires ran along an axis immediately below the spine, reducing the amount of travel for the wires and increasing the flexibility of the snake.

We successfully managed to implement the brute force calculation of servo-space controls from a world-space sinusoid. We also extended our algorithm to allow to the snake to follow a sinusoid projected on any parametrically defined curve, enabling the snake to do things like slither in a gradual curve, or slither along another, larger world-space sine wave.

Mechanical features of the latest snake include a length of over 4 feet, a thermoformed head, improved linkages, and a sturdy yet flexible design. Prior to designing this snake, the mechanical team set down to address all past design issues:

In past designs, the wire linkages would pop out of place during actuation. To address this, we used wire with a smaller diameter, allowing for slippage, and secured both ends of the wire by bending it at ninety degrees to prevent it from easily coming out. Bending the ends an additional 45 degrees allowed for easy insertion. We also attempted to use other linkage designs, made of rounded wire and delrin, but manufacturing time made these designs undesireable.

The SD-21’s mounting rib in prior designs had been cantilevered off the snake. In this revision, we added wheels to its rib. Additionally, to minimize wiring problems, we offset it from the rib allowing us to place wire between.

We noted slight plastic deformation in the Lexan spine in other designs. However, after using the INSTRON tensile test machine in the material science laboratory we determined that it was still the best material for its price range.

In order to reduce the amount of wires, we used 14AWG stranded core cable as a power bus, tapping off 26AWG wires as we needed power. This way, we only had to run one 26AWG wire down the length of the snake for each servo, which made wiring much simpler, and less stiff. We also laser-cut features into the ribs for wire management, making the overall job cleaner and easier to work with.

We added Purple LEDs with a 500 Ohm resistor on each powered rib for underlighting and Amber LEDs with a 500 Ohm resistor in the head of the snake as eyes for aesthetics, and we can’t say that it was a bad decision.