Project Review: 1. Project #3: “Developmentally Learning the Support Affordance of a Platform”“ 2. Should this project be considered for the Best Project Award? No. While the project proposed a developmentally-sound approach to learn support affordances, the experiment and results need a bit more work. 3. Should this project be considered for the top 3 project awards? No, but it is close. 4. On a scale of 1 to 10, how would you rate the overall organization/clarity of the project? The write up is organized fairly well; however, some things are still not clear - specifically, the machine learning problem is not very well defined as far as what the inputs are and what the outputs are. The description of the results only talks in terms of “classification performance” what exactly the robot is classifying isn’t mentioned. Overall, it gets a 6. 5. On a scale of 1 to 10, how would you rate the overall project idea? The project idea overall is very good. It is developmentally sound and it approaches the problem in a novel way. For the idea, this project gets an 8. 6. On a scale of 1 to 10, how would you rate the overall research contribution of the project idea, methodology and/or results? 7.0. Having seen the experiment of this project first hand, I would say that the biggest contribution of this project is the experimental platform, methodology, and collected data. The analysis and the results themselves need much more work in order for this to be publishable. * Overall, is the project report clear, concise, and well-organized? It is well organized, somewhat concise, and clear in some parts. Some parts, however, are not very clear. For example, the description of the classification problem is written in prose - this isn’t literature class. The authors need to use a separate diagram that describes the classification by itself. * How does the project idea and methodology fit within the framework of Developmental Robotics? The methodology first very well. The analysis, not so much. It is also not exactly clear how the labels “control” and “no control” are grounded. * Do the methods, results and contributions of the final project correspond to what was presented in the initial project proposal? About. It was a fairly ambitious project and the results in the final report match the proposal to an extent. * Are there any major details left out with regards to the methods, algorithms, or experimental design described in the report? It is not clear whether the machine learning dataset was balanced or not. If it wasn’t balanced with respect to the class label, reporting the raw % accuracy is misleading. * Do you have any suggestions for improvement and future work? Several: 1. Improve experimental setup so that robot pushes object at arbitrary positions rather than just 3 selected ones. 2. Be more clear about what exactly the robot is predicting, what the inputs are and how the class labels are grounded. 3. Also, the robot should be able to learn more from this data, for example, how do objects with different shapes behave once they’re no longer under the robot’s control. * 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? It has some of the elements in terms of introduction, motivation and related work, but it needs better and more precise problem formulation since at the moment it is not exactly clear what the robot is learning. Proposal Review: 1. Proposal number: 3 “Developmentally Learning the Support Affordance of a Platform”“ 2. Should this proposal be considered for the Best Proposal prize? no 3. On a scale of 1 to 10, how would you rate the overall organization/clarity of the proposal? 8 4. On a scale of 1 to 10, how would you rate the overall project idea? 8 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? The proposal is clear and well organized. * Does the proposal meet the posted proposal guidelines? Yes, for the most part. The related work, however, doesn’t include much work in robotics in which robots have probed objects and observed how they move. * How does the project idea fit within the framework of Developmental Robotics? The project fits very well in framework of development. The method mimics early childhood exploration of objects. * Describe what you like BEST and LEAST about the project idea. My favorite part of the idea is that it deals with learning object affordances. My least favorite part of the project is that the evaluation is not very clear. To me, it seems that there is a lot more that can be learned from the data than just the boundary of support. * Do you have any concerns about the project? Does it seem doable in the remaining time? The three stages of object exploration will probably take quite a bit to implement. The authors might want to focus on just 2 of them and assume the other is known a-prior (e.g., the calibration stage 1 may not be necessary if the project is going to focus on other aspects). * Does it seem too difficult? It doesn’t seem easy. * Are there any major details left out? Yes, some. For example, the evaluation is not very well defined and it is unclear what exactly the robot will learn or anticipate. * Does the idea rely upon technologies that are not currently available? No. * Do you have any suggestions for improvement? Do you have any suggestions for related work that should be cited? I have a few suggestions for the authors. First, it seems that a lot more can be learned from the data other than just the support boundary. The robot could in principle use this setup to learn about the friction of the object and the inclined plane. For example, different pairs of objects and plains will result in different velocities of motion of the object, so perhaps some kind of categorization could be performed. Along with that, varying the angle of the plane could also be used to learn that at different angles, objects move with different velocities. Also, if using object visual features in the model, simply using height and width may be insufficient since the shape of the object also influences how it moves - for example, a ball rolls pretty fast, while a wooden block probably slides slow. Finally, I would also suggest the authors do a thorough related work section so that other work in robotics in that area is accounted for. This way, the project will focus on methods that are novel and ensure that the work can be published.