Sangbae

The paper is easy to follow. But lables of figures are very hard to read and more diagrams would help explanations. The reasoning why they choose body joint position as a control parameter is weak.  More detail result of experiment and comparison with simulation result would make it more convincing. 

In simulation, the "failure criteria" is very unclear. (I think being knematically possible or not is the criterion.) It seems that three wheel-legs (front, middle and rear) are synchronized because of center motor that drives three axes. Their phases are fixed. When it rotates its body, the phase of the center wheel-leg is not mentioned (according to fig. 4, it looks same as rear wheel-leg with respect to the rear body) . The paper does not mention about this in simulation. I'm wondering if the result of simulation would be almost all kinematically succeessful since there is no contact failure by overload. 
Also, in the paragraph says, "... when more than two wheel-legs touch the surface, two feet are chosen to be attached. Thus, the simulated robot fixes two attached feet and drives forward until another foot touches the surface."  this sentence is confusing. 

Fig. 6 shows that the best body joint design only has 50% of success rate in (d) position. which does not seem to be enough for "optimal or good design" . I'm wondering if they get better simulation result if they change other parameter. 

-Sangbae



Daltorio et al:

The authors describe the addition of an articulating body joint to their
whegs robot that uses an adhesive tape for climbing on smooth surfaces.
Not surprisingly, the body joint is invaluable for making transitions
around corners. Similar findings were observed when a body joint
was added to the RiSE robot from Boston Dynamics Inc.

The details of the simulation program are a bit unclear. The authors
state that once a foot makes contact it is not allowed to move tangent
or normal to the surface (in a 2D model ==> it is fixed) except at the
rear leg, which "is permitted to slip 1mm before the front foot reaches
the upper surface." They don't say anything about the contact model
they use for that slip or how it relates to the position of the front
foot. They also don't explain what exactly constitutes "failure" in the
simulation. Bits of text in a few places lead me to believe that if
less than two wheel-legs are in contact with the substrate, or if the
body or middle wheel-leg's spokes contact the substrate they count this
as a failure?

Figure 6 is confusing, probably because of the use of "%" for both the
vert. axis and the different bars. There has to be a better way to
present that data. Also, the data for the first phase (a) could
probably do with more explanation. The authors point to figure 6 to explain
the sharp decrease in figure 5, but the trend for phase (a) is the
opposite.

I am not entirely convinced regarding the authors' treatent of the
magnitude of body joint rotation. Why are large body rotations always
undesirable? The authors use the experimental results from CMWB31as
compared to CMWB00 to justify their assumptions. However, CMWB31 is
unable to make the body joint rotations that their simulations say are
required, given where the body joint is located. The conclusions
regarding the best placement of the body joint would be stronger if the
authors had tested a platform with a non-optimal joint location
(according to their simulation) but adequate range of motion. Then, if
this platform still performed less when than CMWB00 they would have a
convincing argument.