OPW in Review

OPW Round 7 ran from December 11, 2013 to March 11, 2014. During the three months, I’ve been exposed to, in no particular order: automation with bash scripting, SVN with mercurial, C++ in writing Gazebo plugins and worlds, Python with Google AppEngine and CourseBuilder, Amazon’s EC2 via CloudSim and, of course, Google Hangouts. All these came with working on the CloudSim-Ed project, the output of a brainstorming session which combined the proposed a directed, simulation heavy robotics course supplement and the Mentor2 [1] program that targets secondary school students. And while the original simulation challenges were not the same as the envisioned ones, hopefully, the information gathered from this project can be used to launch more focused or creative ideas.

I personally find that simulation could be a very effective learning tool, especially for schools with limited resources. While others may argue that simulation results can never truly match ground truth results, experimentation in simulation will reduce the time tinkering with both the robot hardware platform and the experiment space. This will place more effort in getting more data to improve the software and insight in creating test cases to run with actual hardware. Another point is that we hope that the simplicity of the Web based learning environment will reduce the intimidating and unfamiliar atmosphere that Linux and CLIs can give off. Also, being web based, easy access is already a given.



CloudSim-Ed [2] (working title) is a MOOC developed with Google CourseBuilder [3] with content from Wikipedia [4] and graded challenges made with Gazebo [5] simulations. These Simulations are housed in an Amazon EC2 g2.2xlarge machine instance [6] which is launched through CloudSim [7]. Typically, a student enrolls/registers for the course, goes over the lessons, organized as units and takes simulation tutorials. At the end of each unit, a graded challenge waits. To run the simulations (graded or otherwise) the student must have a set of CloudSim credentials (IP address, Username, Password and Simulation Machine Name) which are used to connect to the machine running Gazebo. To go though the simulation, gzweb is displays the 3D world while an IPython notebook [8] is used to send ROS commands to manipulate the robot’s position.

Several issues were found during development, first on the list was the cost to keep the simulation machine running on the cloud. While a school or an instructor can organize the uptime for these machines, keeping them for the use of a larger internet audience equates to additional expenses. Another issue is with CourseBuilder itself, as the code for assessments can be accessed by anyone – particularly the curious and the clever. CourseBuilder does not seem to have a fix for this as of this writing nor was there an announcement for a fix. Connection issues, scoring and the overall robustness of the simulation challenges are also on the issues list.


Student Reaction

I consulted with a 4th year computer science student with under a year of experience with Linux and ROS and one of his frustrations involved the installation process. Given that Windows was the benchmark experience, setting up ROS in both his Laptop and Desktop and working with Linux was daunting. He, however, warmed up to ROS after exploring and experimentation. The CloudSim-Ed setup seems to have made an impression; noting amazement from the real time robot reaction. The IPython notebook and the view of the code to move the robot did not receive the same reaction since it remained intimidating, especially when asked if his high school self would want to learn robotics with the CloudSim-Ed prototype. Which may indicate that aside from the gzweb + IPython notebook interface, there is a need for a beginner level that uses basic UI, perhaps integrated into the simulation world.


The Experience

Recent news predicts a decrease in the popularity of  Massively Online Open Courses, citing low completion rates (guilty of contributing this) and the uncertainty of their impact to the culture of learn [9] [10] [11]. While this is unfortunate news for this project, the practice of open education itself should not be categorized as a fleeting trend. I would very much like to see projects that embrace free and open education get continued support and feedback from public and private school teachers who use the tools and students looking for additional resources. I am especially rooting for open hardware and getting them integrated into school curricula or special interest clubs for younger students [12] [13] [14]. Having taken part in the OPW with OSRF only solidifies this view. I’m glad and lucky to have taken part in it and I hope it isn’t the last time that I do.