Review for Project 11
Best proposal: No
Organization: 4/10
Idea:         6/10

The goal of this project is to make the robot manipulate a flashlight. To achieve this, the authors want to use the self-detection methodology developed by Stoytchev in his
dissertation [1] and build on top of this capability to learn to control the flashlight to be able to perform a manipulation task in the dark (the authors give an example of
pressing a button).

Overall, the project proposal communicates the ideas adequately. However, the text could certainly be more refined. For instance, this reviewer failed to understand the last
paragraph on Page 3.

One of the main problems of this proposal is the deficiency of the related work -- there is no survey of prior art in Robotics about flashlights. A good place to start is the
Google Scholar search for "robot flashlight". This search reveals references to work by some well-known roboticists [2] [3]. These references could be a good place to look for
more related work. Also, this reviewer suggests a forward search for the literature that cites these references to see if there is anything newer with flashlights.

Again, it was somewhat hard to understand the last paragraph in Section 1.3 -- presumably, the authors implied that a flashlight can be viewed as a tool. In this respect, it is
important to note that unlike the tools used in [Alex's thesis], the flashlight enhances the robot's perceptual, as opposed to motor, skills.

Regarding section 2.1, it is more clear when direct citations from literature are explicitly indicated with quotation marks. This is certainly relevant for the 3rd paragraph in
Section 2.1. The relevant citations are indeed present, but it is not clear that the text is quoted from these citations.

Regarding the four goals, this reviewer believes that it is not realistic to meet all four goals within the time allotted for the project. Since no other criteria for project
success are available, it appears that achieving of the all four goals is meant to be the project success -- and this is clearly too much. This reviewer suggests focusing on goal
1, goal 2, and showing a feasible future approach for goal 3. This reviewer thinks that the bulk of goal 3 and goal 4 should be moved to future work.

Overall, the project idea is certainly interesting. Animals in the wild don't use tools for enhancing perceptual skills like a flashlight or glasses for its intended purpose. To
have a robot really understand what a flashlight and other perceptual tools mean would certainly be a major breakthrough for Robotics.

To conclude, this project is worth the effort, but this reviewer suggests to put more work into the implementation detail, narrow the scope somewhat and perform a more thorough
search for the related work.

[1] A. Stoytchev, “Robot tool behavior: A developmental approach to autonomous tool use,” Ph.D. dissertation, College of Computing, Georgia Institute
of Technology, August 2007.
[2] Kumar et al. Robot and sensor networks for first responders. IEEE Pervasive Computing (2004) pp. 24-33
[3] LaValle et al. An algorithm for searching a polygonal region with a flashlight. Proceedings of the sixteenth annual symposium on Computational geometry (2000) pp. 260-269