Automated Theme Allotment to Optimise Learning Outcomes in Robotic Competition

In the e-Yantra Robotics Competition we teach robotics through a competition paradigm. Here, 'Themes' or problem statements, are allotted to participants who clear a preliminary selection test. Theme allotment to student teams is a challenge and is time-consuming since we need to fit a Theme to a team capable of addressing the challenges denoted by a Theme. This allotment of Themes considers multiple factors such as team members' experience, expertise, Theme complexities, domain requirements and other constraints set on a Theme by the respective Theme designers. Various machine learning techniques were analyzed and experimented with; however, the problem of project tagging could not be solved by machine learning algorithms alone due to certain randomness in data and human bias involved in previous Theme allotment techniques. We thus devised a new metric that projected multiple parameters onto three dimensions, namely Algorithm, Mechanical and Electronics. Further clustering and analysis were done based on these three dimensions. The predicted clusters were subjected to conditional random sampling based on Theme constraints for actual project allocation to ensure an unbiased and a better allotment of Themes. This led to allotment of Themes such that it provided control to Theme designers, whilst automating the time-consuming process.