1. 4 : "Learned Shape Affordances through Simple Tool Use"

2. no

3. yes

4. 8 Overall the report was fairly clear and well-organized, but 
	 grammatical errors and other typos detracted from it.
	 
5. 10 This idea has a lot of potential

6. 5 While the project idea has a lot of potential, the team was
	 bogged down by a lot of implementation details that prevented
	 them from really getting good results.


The project report is very clear and well-organized, though at times
details are rehashed, making it not quite as concise as it could be.
The main problem with the report itself is the prevelence of typos. It
is evident that much of the proposal was recylced to create the final
report (not a bad thing). There are many places that use words such
as "will" instead of "did." As well there are also a lot of
grammatical errors.

This project fits very well in the framework of developmental
robotics. It is a direct extension of the teaching professor's own
PhD work. The project deals with learning the affordances of a small
set of objects when acted upon by a tool. The related work does a
good job of firmly grounding this project in the field.

The best part of the project is how cool the project idea is. Since
it is a direct extension of Alex's work, it has a lot of potential to
produce good results. This project could easily be the first step to
learning more complex tasks with tools.

The worst part of the project was the implementation details. As I
said in my initial review, this project was very ambitious. Due to
this, the team ran into a lot of issues with things like getting the
simulator to work, adjusting the parameters correctly, and the like.
This detracted a lot from the project and prevented the team from
getting really good results.

The methods, results, and contributions correspond fairly well to
what was initially proposed. Due to the over ambitiousness of the
project, though, the last two iterations of the development cycle had
to be cut out. This was not unexpected though. Of the parts of the
project that were done, they fit fairly well with what was propsed.

This paper covers everything in very fine grained detail. Everything
from a long description of the motivation for the project to details
on the debugging process is included. This is most likely due to the
25 page minimum on paper length. Anything that could possibly be
deemed relevant to this project is included, which helps
understanding of the project development, though it does seem a bit
cluttered at times.

The experimental results do not demonstrate success. While they did
eventually manage to get 100% success rate with the cylinder, and a 66%
success rate with the square, the only got 8.33% success rate with
the triangle and a 0% success rate with the rectangle. These success
rates are not very good. This is due in major part to the large scale
of the project and assumptions that had to be made along the way, as
detailed in the section "Lessons Learned."

To improve this work, the main thing that can be done is to scale
back the scope of the project. The team only had a month and a half
to implement the entire thing, making this kind of project very hard
to complete. Also, they may have been able to use a pre-built robot
simulator as that seemed to be the root of most of their issues, or
they could have used the actual robot, which is easier to do than
building a simulator.

This project in its current form is not publishable. The idea is
there, and if its actually carried out this could be a very good
paper. Currently, the results aren't good enough, as well as the
exact methodology isn't robust enough (e.g. only pushing in the four
cardinal directions instead of an arbitrary direction). If this
project were fixed up, then it could make a very good conference
paper.