Recently, there has been interest in low-power robots that are either very tiny in size or can operate on long duration missions; it is necessary for the algorithms onboard these robots to also be extremely energy efficient due to low energy constraints. The goal of our research is to develop active learning algorithms to decide when to spend extra energy to train a DNN used for navigation in order to get better predictions in the future. This research would enable resource-constrained robots that can intelligently decide whether it is worth it to spend extra energy on retraining in order to navigate in novel environments different from what it has been trained on, enabling applications for miniature and long-duration robotics.

I am participating in SuperUROP to further pursue my interests in robotics and machine learning and apply what I’ve learned in classes I’ve taken. I also hope to improve my research and presentation skills.