1. Proposal number: 4 2. Should this proposal be considered for the Best Proposal prize? (yes/no) no 3. On a scale of 1 to 10, how would you rate the overall organization/clarity of the proposal? (1-10) 6 4. On a scale of 1 to 10, how would you rate the overall project idea? (1-10) 7 Then write 2-3 pages of helpful feedback to the proposal's author(s). The following questions should help you organize your feedback: * Overall, is the proposal clear, concise, and well-organized? * Does the proposal meet the posted proposal guidelines? * How does the project idea fit within the framework of Developmental Robotics? * Describe what you like BEST about the project idea. * Describe what you like LEAST about the project idea. * Do you have any concerns about the project? * Does it seem doable in the remaining time? * Does it seem too difficult? * Are there any major details left out? * Does the idea rely upon technologies that are not currently available? * Do you have any suggestions for improvement? * Do you have any suggestions for related work that should be cited? * Any other comments or suggestions? OVERALL: The main goal of this paper is to show how a simulated robot can learn to move a specific-shaped puck over a similar-shaped area. The robot will learn by performing random babbling behaviors using an L-shaped stick. This is an extension of learning the affordance of specific tools. In this experiment, the robot is learning the affordances of differently-shaped pucks. A robot that learns something differently for each shape would probably perform a unique set of behaviors for each shape in order to maneuver the puck onto the goal area. The proposed idea is good. It has not been done before. The ideas in the proposal seem like they will be effective toward enabling a developmental learning robot to learn about the shape affordances of pucks. While the presentation of these ideas is fairly clear, there are several ways you could improve the communication of this work. The first few pages of the proposal are fluffed up and should be condensed. The purpose of this project is unknown until the approach section on page 5. Consider merging the 'Motivation,' 'Demonstrated Need,' and 'Approach' sections. What I like most about this work is that you are studying shapes. Your own literature review shows that there's not a lot of research in this area. The study of shapes is important. Differently shaped pucks can move in very different ways. Robots can learn how they are different and possibly apply their knowledge to novel shapes (an idea for future work). The project seems doable in the remaining time. The team identified the hurdles of the project. Bullet and OpenCV are the main challenges that need to be overcome. I know that the challenges presented by OpenCV can probably be overcome due to my own experience with OpenCV. IMPLEMENTATION DETAILS/QUESTIONS: --Why would you want to identify the orientation of the differently-shaped objects? --Would it be possible to include more pucks in order to get statistically significant results? --It looks like you're identifying the puck by its center pixel position relative to the position of the yellow marker. Could it be possible for the robot to confuse two differently-shaped pucks, e.g., the circle and the triangle because the distance between the two points is the same? To show that the robot can learn about shape affordances in a general way, the algorithm should work for many different pucks that have the same distances between the center point and the marker point. --The obj_act data structure seems insufficient for capturing all of the data from the robot's behaviors. The first four variables describe specific parameters for a single behavior. The last two variables can only be attained after the robot has performed all of its behaviors. --With so many different parameters for the push behavior, it looks like you will need many interaction trials. How many interaction trials do you think will be sufficient for learning the shape affordances of each object? Will a two minute duration of interaction be long enough to learn about each object? How many behaviors can the robot perform in this amount of time? --Would you consider evaluating what the simulated learned robot using novel shaped pucks? It might be possible for the robot to integrate its knowledge from the other pucks to manage the new object. SPECIFIC COMMENTS ABOUT EACH SECTION: TITLE: What is a "shape affordance"? INTRODUCTION: Based on the title of the paper, this paper is about learning about shapes and how different shapes behave when the robot interacts with them. The robot will learn about how the shapes of objects affect the outcomes once the robot interacts with the objects through tool-using behaviors. The introduction is focused more on the tool-use aspect of the project rather than the shape-affordance aspect of the project. You could clarify the point of this paper by reconfiguring this paragraph. MOTIVATION: The ideas in the first paragraph are good. The structure of each paragraph makes sense. The ideas can be clarified by iterating through a few more drafts. Consider a few more for the final paper. The focus on the ideas in this paper are lost in the third paragraph. Consider removing those sentences and/or reorganizing the structure of the motivation. The text states clearly that existing work has studied a robot's affordances when interacting with a variety of objects, and that more work is needed to address how a robot's affordances change as the object being interacted with changes. It is not clear, however, what is the advantage of learning about a "shape affordance." This is the central theme of the paper and was not motivated. It is hard to believe that one of the contributions of this project is that it will help to "gain deeper insight into developmental psychology of humans and animals." No supporting evidence was provided to back up this claim. How can your study can achieve this? Is that the primary motivation for this study? Is there a shortage of information about shape information in the developmental psychology literature, which only a robotics study can provide? AUDIENCE: What specific automated processing industry should take notice of this study? DEMONSTRATED NEED: Find a positive angle, which conveys to the audience that you're building on previous work. Instead of your "tool use has not been well addressed..", it would be better to say "Tool use has remained a primary task in the field of robotics since its inception. Traditional methods of tool use are in the form of an engineered solution to a factory assembly-line automation problem. Recent work has suggested that some robots can learn about the affordances of tool use, and then apply this knowledge to real-world problems in a more general way than traditional approaches." Or you could say, "the study of tool affordances is a fast growing area in the field of developmental robotics." This section reads as motivation. Remove it by merging it with the motivation section. (And re-organize your motivation) PREVIOUS WORK/RELATED WORK: The ideas about the affordances are good. Consider adding some work by your [8] reference to this discussion. That paper introduced the term "affordance" to the field of robotics. Also, if you have time, read J.J.Gibson's book on affordances and cite that instead of the robotics studies. It would be beneficial to motivate this work if you could find a point in gibson's book or somewhere else on "shape affordances." The most relevant citation you have included (besides alex's previous work) is that by ugur et al, since their robot learned how traversability is affected based on the shape of the objects. The related work should mostly focus on autonomous learning about the shapes of objects. So, try to find more like that. The Sinapov and Stoytchev reference is unrelated. PREVIOUS EXPERIENCE: You sound qualified to do this research. APPROACH: Move this to the motivation section. The reader is dying to know what you actually did, and this is the first place you say it. EQUIPMENT: Are you using the same simulator that stoytchev used in his work, or are you writing your own? DEVELOPMENTAL MODEL---TIMELINE: The headings don't match. In the developmental model they're called iterations. In the timeline they're called phases. EVALUATION MODEL: Figure 10 shows a star, but this is not in your list of objects in the equipment section. Consistency makes it easier to understand your project. ß REFERENCES: Sometimes the quality of a paper can be determined by how much attention to detail was put into the references section. The references have typos and are missing information. This reflects poorly on the quality of the proposal. Consider fixing this for the final project. Some information that is consistently missing from your references includes page numbers, journal article, volume, number, and year of publication.