Smith, Joshua RBoots, ByronYang, Boling2024-02-122024-02-122024-02-122023Yang_washington_0250E_26427.pdfhttp://hdl.handle.net/1773/51133Thesis (Ph.D.)--University of Washington, 2023We explore the intersection of robotics and physical games, introducing innovative methods to enhance robot capabilities in real-world scenarios. Recognizing games as potent tools for refining cognitive and sensory-motor skills, we leverage diverse perspectives, including benchmarking, human-robot interaction, and robot learning. Through these lenses, we demonstrate that physical games provide an optimal environment for robots to hone their problem-solving and manipulation capabilities. Our research delves into how puzzles, exemplified by the Rubik's Cube, can bolster robots' sensing and manipulation skills. Further, we delve into the development of human-robot interactions in competitive exercises, showcasing the potential benefits of a robot as a contender. We also present a game-theoretic automatic curriculum learning algorithm, aiming to enhance the learning efficiency of robots in competitive gaming contexts. Finally, we advocate for the application of game-based methods to real-world robotic tasks, particularly object handling and rearranging within warehouse storage units.application/pdfen-USCC BY-NC-SACompetitive Human-robot InteractionGamesMachine LearningRobot ManipulationRoboticsComputer scienceRoboticsComputer science and engineeringThesis