Lifelong Learning from Videos and Physical Interactions
Abstract
Statement of Work:Purchase a GPU cluster and mobile robots with grippers for research in object recognition and manipulation and scene understanding by robots.Objective:Purchase a GPU cluster and mobile robots to perform research in learning to reason from visual data for objectrecognition and scene understanding, and in learning object manipulation by robots, as well as integrating vision and manipulation.Approach:The proposed equipment will support a number of DOD research projects including the ONR MURI visual common sense, Army Robotics CTA, , DARPA Memex Program, as well as IARPA programs. The equipment, a GPU cluster and mobile robots with arms and grippers, will support cutting edge research in learning object manipulation by robots and learning to reason from visual data for object recognition and scene understanding, as well as integrating vision and manipulation. Current robots (autonomous agents) have limited capabilities in interpreting images/video (for objectrecognition, scene and event understanding) and in manipulating objects. The research supported by the proposed equipment (a 100 GPU cluster, as well as a mobile robot), takes a novel and effective approach for developing these capabilities based on advanced machine learning methods from observing and manipulating a huge number of objects and scenes, as the robot explores its environment with minimal human supervision.Overall Merit and ONR Mission/Relevance:The proposed equipment will support a number of ONR/DoD funded projects in robotics and image/videounderstanding. The proposed equipment will support research in field related to the ONR Focus Area of Autonomy. The proposed equipment will further enhance the research and education programs at CMU in developing capable robots that are of importance to the Naval/DOD missions. The research supported by the proposed equipment is ambitious and cutting edge, and important in Naval/DoD missions. These projects have the potential for significant advances in developing capable robots. The PI and CMU are leaders in robotics and image understanding.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 30, 2016
- Source ID
- N000141612905
Entities
People
- Abhinav Gupta
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy