Somesh Daga

Master's Student

Contact information


Location: E7-3444

Link to personal website or CV



BASc in Engineering Physics (2017) from the University of British Columbia (UBC)

Current research topic

Deep Reinforcement Learning for Robotic Manipulation

Description of research topic

Robotic Manipulation tasks like Reaching and Pick-and-Place are an area of significant interest in the deep reinforcement learning (RL) space with widespread practical applications. My current research is focused towards evaluating the effectiveness of state-of-the-art Deep RL methods with non-linear function approximators on such continuous control tasks using robotic manipulators. Subsequently, I aim to address issues commonly associated with Deep RL methods in this space e.g sample complexity, generalizability, training time, performance etc.

Research interests

Reinforcement Learning, Path Planning and Obstacle Avoidance for mobile robots, Simultaneous Localization and Mapping (SLAM)