REVIEW GUIDELINES Each review should start with the following information 1. Project number and Title Project Number: 4. Title: Learned Shape Affordances through Simple Tool Use 2. Should this project be considered for the Best Project award? (yes/no) no 3. Should this project be considered for the top 3 project awards? (yes/no) yes 4. On a scale of 1 to 10, how would you rate the overall organization/clarity of the project report? (1-10) 9 5. On a scale of 1 to 10, how would you rate the overall project idea? (1-10) 7 6. On a scale of 1 to 10, how would you rate the overall research contribution of the project idea, methodology and/or results?  (1-10) 8 Then write approximately 2 pages of helpful feedback to the project's author(s). The following questions should help you organize your feedback: Your final project report was evaluated as if it were submitted for publication in a workshop or a conference. The comments below would make it more publishable. Introduction --------------- Instead of giving the title of Alex's paper, you should reference a citation. Motivation ------------- There are no citations in your motivation section. Usually, research is motivated by identifying holes in existing work. Because you have not cited similar work, it is not clear what your contribution is to the field. At this point in your report, it is clear why robots should be able to manipulate objects that have different shapes, but it is not clear that what you said about existing research is actually true. The introduction has provided evidence that Alex's work would have been more thorough if differently shaped blocks were used, but it is not clear from either section that his work should have been evaluated by varying this extra detail. Here are some examples of where you should have had citations: ---"Current robotic systems are programmed to work efficiently, but almost all are quite task specific and are incapable of being extended to work on other tasks without reprogramming." ---"A robot may win at chess, but that is a testament to its data processing, memory and efficient search and retrieval routines than actual IQ " In the paragraph that starts "thirdly,…", you make the conclusion that a robot should learn affordances rather than be programmed with affordance knowledge. The paragraph does not provide supporting evidence for this conclusion. Instead of talking about affordances, you might describe, in general, the five principles of developmental robotics (one of Alex's paper that we read in the first week of class), and the verification principle, in particular. You briefly imply that different shapes have different affordances in the last paragraph of the motivation section. It would be nice if you gave an example of how the robot is afforded actions by one shape that are different from those afforded by another shape. For example, you could say that a round shape affords rolling, a flat shape affords staying in place, and a triangular shape affords paper football. Audience ------------ How can people in the automation industry benefit from your research. I can't immediately think of good application for your work. Demonstrated Need ------------------------ The demonstrated need for tool use is clear. What is not clear is the need for knowing how to move around objects that have different shapes. You could cut some of these details from the motivation section in order to fix this one. Related Work ----------------- The related work section is thorough. You cite a variety of literature for each section. The section could be improved by condensing the discussion of affordances. You might choose to focus on a single source for the definition of an affordance. You might also limit your discussion of affordances to those points which have something to do with the shape of an object. You quote Gibson directly, but fail to mention the source of the quotation. Your motivation would be stronger if you cited the related work that you've identified. Approach ------------ This reads like an abstract. Make it the abstract. Virtual Objects ------------------ The way your robot used the hooked-stick tool is different from the way the robot in Alex's work used the hooked-stick tool. You should clearly state this difference since your research is a direct extension of his work. Algorithms and Data Structures -------------------------------------- Many of the sentences in this section were written as if you just finished coding a bunch of 'if' statements. This text could be improved. The specifics of your implementation are not nearly as good as they could have been, given a few tweaks. The behavior applied to a puck is one that was already performed by the robot in the behavior babbling stage. You could easily modify the algorithm to select a few good candidate behaviors and then apply linear regression (i.e., select three babbling behaviors that produced results like the ones you need and then calculate the new behavior using a weighted linear combination of the three). You probably could have gotten away with much fewer babbling trials and would have definitely gotten much better results. This simple fix would have made your results waaaay better, maybe even publishable (you could still try it…). Implementation Methodology ----------------------------------- This section should be removed from the paper. I don't need to know what data structures you used. I don't need to know the implementation methodology. You guys made a website? What for? Evaluation Methodology ----------------------------- Some of the information here is unnecessary as well. Results --------- Figure 15 shows that there's room for improvement in the learning methodology. A robot shouldn't have to push a block 500 times to get it into the right position and orientation. Conclusion ------------- The conclusion is accurate. Lessons Learned -------------------- Remove this section from your paper. * Overall, is the project report clear, concise, and well-organized? Your report had at least 10 pages of fluff. I would have given you a higher score if it didn't have this stuff. * How does the project idea and methodology fit within the framework of Developmental Robotics? This is an extension of one of Alex's papers. * Describe what you like BEST about the project? You have done enough work that, if you do just a little bit more, you might possibly be able to get a publication out of it. * Describe what you like LEAST about the project? The method you used for choosing a behavior is too simple. The results captured this. See what I wrote above, which starts with "The specifics of your implementation …" * Do the methods, results and contributions of the final project correspond to what was presented in the initial project proposal? Yes. * Are there any major details left out with regards to the methods, algorithms, or experimental design described in the report? No. * Do the experimental results reported in the paper demonstrate success? Somewhat. * Do you have any suggestions for improvement and future work? Yes. See the feedback section. * How close is the final project report to being publishable as a conference or journal paper (consider the research papers that were part of the course reading)? What would it take to get there? Maybe close. You would need better motivation for shape learning. You also need a better learning algorithm. I suggested a better algorithm for you in the feedback section.