Robust drone navigation in environments with adverse conditions such as poor lighting is an unsolved problem in autonomous navigation with important applications in search and rescue operations and exploration of dangerous environments. For example, state of the art simultaneous localization and mapping (SLAM) algorithms like ORB-SLAM 3 fail to detect salient features in the presence of moderately heavy fog. The fact that unavoidable visual impediments lead to localization loss and our observation that humans tend to use multiple sensory modes to navigate in the dark motivates looking at multimodal localization methods. Thus, we propose a novel, reinforcement learning-based algorithm for haptic localization, mapping, and relocalization for drones in vision denied environments.

I’m participating in SuperUROP to gain research experience in robotic navigation, a field I’m excited about. I hope that the structure of SuperUROP will help me to streamline the research process and contribute to the field of robotics.