Introduction to Machine Learning in Undergraduate Curriculum using LEGOLAS: A LEGO Based Low-Cost Autonomous Scientist

Cutting edge topics such as machine learning are frequently not covered in the undergraduate physics curriculum. We decided to introduce this topic to students through a project-based class, where students constructed a LEGO-based, low-cost autonomous scientist (LEGOLAS). It is a robot that uses machine learning techniques to autonomously derive the Henderson-Hasselbalch equation, reducing the tedious experimentation and calculations typically involved in acid-base experiments. Originally, it was developed via collaboration between NIST and UMD scientists as a teaching tool to introduce students to machine learning techniques and algorithms [1].

The robot was built by students from scratch. Funding for parts and electronics came from the SPS National Research Grant. The project took two semesters to complete. Students really enjoyed working on it and, along the way, learned about 3D printing; electronics such as Raspberry Pi’s, Arduino computers, and pH sensors; and other topics and skills that are usually not taught in undergraduate curriculum. Machine Learning was introduced through Bayesian statistics and Gaussian process regression, which LEGOLAS used to derive the Henderson-Hasselbalch equation in our experiment. We plan to use this device to develop other autonomous experiments and for departmental recruitment and outreach. In this talk, we will share our experiences and lessons we have learned while constructing, calibrating, troubleshooting, and using LEGOLAS.