Snaptoe001 development documentation
October
9 to
Motohide Hatanaka
Abstract:
Compliant toe mechanism for a
gecko-inspired climbing robot was developed. The toe deploys segment by
segment
from a curled up state. There are angular interlocking mechanisms
between
adjacent segments that are released only when the proximal segment has
been
pressed sufficiently for its adhesive material to function after which
the
distal adjacent segment can move. The toes are locked again when curled
back. The
entire development process including need identification, concept
development,
and fabrication is recorded. The design seems promising although there
were
flaws in fabrication. A design methodology named scenario alteration
and a
design philosophy named sharing and distribution are also mentioned.
Keywords:
Scenario alteration. Sharing and distribution.
Aim/requirements: “Conform + Press + Peel off”
A wall climbing robot that relies on a gecko-skin-like dry adhesive material requires a conformal support structure for the material and the application of appropriate forces to maximize its contact area with the climbing surface. The material also needs to be peeled off after each step. The Snaptoe is intended to provide the skeletal conformability of the toes and the forces needed for adhesion and release.
Inspiration: Gecko toe
The conformability observed in a gecko is provided by the skeletal deformability of the toes as well as the compliance of its flesh and then finally by that of the fractal structure of the skin itself. The flesh and the microscopic features on the skin require some force to press them against the climbing surface so that it can passively conform to it. Later on, gecko toes are curled away from the surface to release adhesion.
Scenario:
The basic function scenario is to conform, press, and then peel off as stated in the aim/requirements section. However, the order of events can be altered slightly. Notice that the conformation is an event that occurs over an area which can be segmented into smaller areas which in this case would be the toe segments. So is pressing. Hence, there are three scenarios available for adhesion; (A1) fully conforming and then applying pressure at once to the entire length of the toe, (A2) fully conforming but then applying pressure segment by segment, or (A3) conforming and applying pressure sequentially, segment by segment. The peel off scenario similarly has two possibilities; (B1) applying equal peel-off force to all segments so as to detach all at once or (B2) applying force segment by segment sequentially from the most distal segment so as to detach segment by segment. The combination of scenarios (A1) and (B1) can be realized by a mechanism developed by Shigeo Hirose (add reference) but this scenario has some disadvantages over others. First of all, the design by Hirose seemed too complex to be fabricated in the desired size of about a centimeter per segment. Having an actuator per segment would also work but this solution would also be too bulky. The author has done some literature search on simpler design alternatives but has not been able to find any. Moreover, applying force at once, either for adhesion or for peel off, will require a stronger actuator compared to when forces are applied segment by segment. Hence, the design activity was carried on to realize the combination of (A2) or (A3) with (B2).
Scenario A1: conforming and then pressing at once
Scenario A3: conforming and applying pressure segment by
segment
Key design methodology and philosophy:
Scenario alteration is a method of searching for alternative scenarios after the conception of an initial scenario. Reordering the existing events is the first and the most obvious thing to consider. Some event orders are interchangeable, some are not. Furthermore, as is demonstrated in the paragraph above, one can check if an event can be broken up into smaller pieces and dispersed. Event segmentation is possible when the result is a continuous change (e.g. gradual increase in contact area) rather than a discrete change (e.g. digital signal change between 0 and 1). It is also possible when the event is a set of non-interdependent events, discrete or continuous. The opposite of event segmentation is to unify discrete events. In theory, this is just a special case that could result after the event reordering operation.
Sharing and distribution is a design philosophy that allows the designer to achieve a requirement via efficient use of resources. For example, in this Snaptoe design the actuator output is shared by the segments by distributing the events over time. The actuator size and power requirement are reduced as a result saving weight and size of the system. Sharing and distribution do not always coexist in a design and they can even be contrary concepts. For example, four wheels of a car can be powered by four distributed motors rather than sharing the output from a single engine.
Concept development:
There was an earlier version of the toe that was driven by a pair of tendon cables anchored to the most distal segment. This mechanism deformed like a single-support beam with point load at the distal end and could not deliver much force to the other segments. It also had issues with conforming to uneven surface geometries because of the loading condition.
Previous
version of the tendon-actuated toe
Conformability and loading evenness can be improved by incorporating multiple actuators or by adopting the design by Hirose. However, both of these solutions seemed difficult to realize in the desired scale. Instead, sequential force application from the proximal to the distal segments was conceived as an alternative solution (scenario A3).
The first idea was to have clutches in each of the segments that release the tendon cable only while the segment is pressing the climbing surface harder than a certain threshold force. The force from the cable would be applied mainly to the most proximal segment with an engaged clutch. Several spring-loaded clutch mechanisms were considered. However, it was realized that the proximal clutches would keep engaging and not allow distal segments to function.
Spring-loaded clutching scenario
The solution to this
problem was to
have bi-stable clutches that would
stay open once the threshold pressure had been reached. A pair of
parallel
buckling leaf-springs was thought to be appropriate for the purpose.
This
system also required a re-engagement mechanism for the clutches after
curling
the toe back. One of the two main solutions was to have a fixed
structure for
resetting that the segments would interact with at the curled-back
stage. This
concept was aborted due to the difficulty of applying force evenly on
the
segments. The other was to have beams stretching out from the segment
proximally adjacent to the segment of interest that reaches into the
segment to
re-engage. The geometrical consideration of the beam also led to an
idea that
the beam could function as a sliding rotary bearing between segments.
However,
the difficulty of having pre-loaded leaf springs in the small space
discouraged
the further pursuance of this concept. One of the main difficulties was
to
ensure reliable cable securing force while maintaining the leaf springs
flexible enough to buckle open at the threshold pressure.
Buckling clutch scenario
An alternative to using leaf springs was to use magnets. When magnetic forces are used at the both ends of the clutch actuation, the activation force profile is quite similar to that of a leaf spring clutch. This seemed to be a more reliable idea than the previous one except that there was a slight concern about the magnets interacting with ferrous metals outside the toe.
Soon
after, a
redundancy in the
concept was noticed while finalizing this magnetic clutch design. The
clutch
has the ability to engage anywhere along the length of the tendon cable
while
it only needs to do at the state when the toe is fully curled back.
This was
considered to be an over specification which could be simplified. The
new
concept was to rigidly interlock the
adjacent segments until when the proximal segment is pressed against a
surface
over a threshold force to release the lock. This would also
eliminate both
the concern of cable wear and the need to fine tune the mechanism to
catch the
cable. It took the author four weeks to reach this final concept
including one
week of interruption to develop a robot foot.
Design constraints:
Dimensions were finalized to satisfy the segment pitch of 20mm and deflection of ±60° per joint.
Fabrication planning and results:
Shape deposition manufacturing (SDM) process planning was done by drawing a side view schematic of the part and labeling the material addition order by two strategies; (1) pouring as early as possible and (2) postpone pouring as much as possible. The first approach was taken in the actual fabrication although there was no clear advantage or disadvantage in either one. Another planning strategy which would have been useful is to pour in layers based on the depth such that the materials at equal depth in the mold would be poured simultaneously. This is because some difficulties including end mill fracture and bubble capture in poured plastic was experienced due to the molds with unusually high depthwise aspect ratio.
The minimum mold wall thickness was 200µm which is a thickness that had proven to work well in previous applications. However, material fusion across the wall which is suspected to be the result of wall melting has been experienced. This might be due to the fact that low-temperature melting wax was used whereas the previous experiences have been with high-temperature machining wax.
One of the two holes for tendon cables was fabricated by using sacrificial wax which was to be melted out after extracting the part from the mold. However, many of these holes were filled with plastic. Possible causes include air trapping in the narrow cavities (approximately 70mm long × 1mm wide × 4.5mm deep) due to wax not flowing in deep enough before solidifying and wax melting after plastic pour. The former cause is thought to be more likely and it can be prevented by avoiding high aspect ratio cavities or by assuring the elimination of air bubbles by using a thin soldering iron tip or something similar to melt the waxes deep inside the mold cavity.
The other hole for the tendon cable was prepared by embedding a thin music wire (approximately f1mm) which was later removed. The music wire was either dip coated in wax or spray coated with powdery Teflon. The former proved to be easier to remove. Furthermore, having one long wire go through all of the segments was easier for both preparation and removal. However, the wax dip coating requires some skill. It would be worth trying some other kind of spray coating which leaves a thin film rather than powder. It is also important to remember to leave some extra length at the end of the wire to be held with pliers for pulling out. An end mill was broken while trying to machine just over the embedded wire which happened to be slightly higher than planned and interfered with the tool. Such an accident can be avoided by machining not too close to a hard embedded part or by using a more machinable material for embedding. For example, embedding a Teflon wire would eliminate the concern for tool damage as well as the need for coating.
Extracting the part from the mold without breaking its fine features was the most unexpected difficulty experienced in this fabrication because it was unconsidered. Previously fabricated SDM parts were never as fragile as this one and did not require much attention for removal. The main difficulty was involved in breaking off parts of the mold (e.g. thin walls) with the finished part which often required forces unbearable for the surrounding fine features. This problem can be improved by using a removable bracket for the mold or by applying mold release at appropriate locations between the bulk wax block and the fine mold features like the thin walls.
The
fabrication process took eight to ten working days. It could have been
finished
in about six days if needed. Tool breaking was one of the reasons that
slowed
the process down.
Operation results:
The toe cannot be deployed with
a cable due to manufacturing errors. Pressure pads do not work either
for the
same reason. However, it can be curled back with cable actuation and
the
interlocking mechanism works very smoothly.
Future work:
The concept seems to be promising. It needs some modification in the fabrication process to improve its reliability. Further scale down would be ideal (e.g. to 10mm segment pitch rather than 20mm) and that may require the incorporation of stronger materials to prevent failure.
The scenario alteration method and the sharing and distribution concept are to be explored further theoretically to investigate their broader application. Scenario alteration shall be helpful for inspiring design alternatives for designers that cannot find a solution constrained in a certain scenario. Sharing and distribution can help eliminate excess in a design.
Conclusion:
A proof-of-concept prototype for a deploying toe mechanism for a gecko-inspired climbing robot was developed. A solution different from the first conception was derived by considering alternative scenarios of the mechanism in action. This process was named the scenario alteration method and is to be further developed. The final concept of interlocking segments named Snaptoe001 was reached after considering various cable-clutching mechanisms. The design incorporates the philosophy of sharing and distribution which helps to make efficient use of parts in a system. The final product had issues with the fabrication processes but the problems have been identified and improvements have been suggested. The next steps are to validate those solutions and also make similar systems in a smaller scale.
Acknowledgement:
The author would like to thank DARPA for
sponsoring this
project, Stanford Rapid Prototyping Lab for providing the fabrication
facilities, and the geckoes for inspiration.