1. Project number and Title

Project number: 2
Title: Behavior-Grounded Object Identification, Grouping and Ordering by a Humanoid Robot

2. Should this project be considered for the Best Project award? (yes/no)
yes.

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)
8

5. On a scale of 1 to 10, how would you rate the overall project idea? (1-10)
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?  (1-10)

7

Then write approximately 2 pages of helpful feedback to the project's
author(s). The following questions should help you organize your
feedback:

* Overall, is the project report clear, concise, and well-organized?

Yes.

* How does the project idea and methodology fit within the framework
of Developmental Robotics?

The project idea fits within the framework of Developmental Robotics very well. Developmental Robotics assumes that the robot should be like human to undergo a development. During the development, the robot will actively explore the environment around, to build its own representation to the environment or understand the environment on his own way as well as improve its skills for doing some tasks in the environment, so that it can adapt the environment better. 

In this project, the robot will be programmed to use several different explotorarybehaviors say lifting objects, pushing objects, poking objects, and so on.During the exploration, multimodal sensor data will be recorded as well as generalized to represent the properties of objects showed during the exploration. Robot's own representation for the objects will be built by its own exploration and will be used for it to solve the tasks of object identification, sorting as well as relational learning. Some unsupervised learning algorithms make sure that the robot can learn on its own to go through the developmental processes on his own.

* Describe what you like BEST about the project?

I like the relational learning best in this project. In the relational learning, two objects will be compared in their weight or size.  I saw few papers for solving this kind of problem, but I think this is also very important problems for the future robot. For example, selecting suitable tools will need the relational learning.

* Describe what you like LEAST about the project?

N/A

* Do the methods, results and contributions of the final project
correspond to what was presented in the initial project proposal?

The methods, results and contributions of the final project correspond to what was presented in the initial project proposal very well. 


Proprioceptive featrue is extracted using n-bin average method.Auditory feature is extracted by first getting the Discrete Fourier Transform (DFT) and then transforming the DFT into a 2-D histogram across time and frequency. The visual feature is represented by color, aspect and the size. Finally, the hand proprioceptive data is represented by the potions of the motors controlling the finger. k-NN is used to do the learning work. The confusion matrix is used to evaluate the recognition results. ISOMAP is used to group the object into a category. The object ordering task is also achieved well.  

* 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?

Yes, the recognition precision is high enough to be demonstrated successful. 

* Do you have any suggestions for improvement and future work?

The ordering task may become more meaningful if the ordering is for choosing a suitable object. To decide if an object is suitable for a task, the criteria will not only the weight or the height, but maybe the combination of several properties. So, how to order the object based on multiple properties may be a possible future work.

* 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?

There are some new stuff in the project, which may be taken to get into a conference or journal paper. First, more sensory modalities like vision and the hand proprioception are added to get more properties of the object and to possibly boost the recognition result. Second, more behaviors are also performed to get  more properties of the object. Finally, the task of ordering object solved in the project seems to be novel in robotics.