College of Science & Engineering
The Robotics: Perception and Manipulation Lab is actively engaged in research related to robot learning and control, focusing on perception and manipulation. This involves various aspects such as deep learning applied to manipulation, teleoperation, tactile sensing, robot control, and tasks related to mobile manipulation and navigation.
The group's research uses MSI resources in two major ways. First, there is a significant need for computational power to facilitate machine learning and artificial intelligence research. This includes training models on extensive datasets comprising text, images, and sensor data inputs, enabling the interaction with established policies. Second, the data generated in this research is substantial. The researchers gather various types of robot data, encompassing joint angles, end-effector positions, mapping information, images, and more. These data streams are high-frequency, exceeding rates of 30Hz and sometimes reaching 100Hz, often spanning across minutes. The process of loading this data is time-consuming due to its volume and size.