Distributed Deep Learning Mobile Sensor System
Abstract
A testbed for integrated, systems level testing of distributed mobile ground and aerial sensors, capable of distributed and centralized learning, is proposed. The testbed will consist of 10 mobile ground sensors, 3 aerial sensors, and an o???-board GPU system. Each mobile sensor includes a stereo camera and GPU for learning onboard; the ground mobile sensors will be further augmented with lidar. The o???-board GPU system is a high end deep learning system capable of large, fast batch learning. The facility will enable a wide variety of research validation experiments to be realized, including distributed sensing and surveillance, distributed and centralized deep learning, cooperative information planning, and tight coupling between sensing, perception, planning and learning. The number and type of sensors will also enable studies in larger scale networks (ad hoc, tree, and other topologies), and hierarchical architectures. The facility will be used support two ongoing ONR research programs, including an ONR Basic Research Challenge (BRC) grant focused on using deep learning and control concepts for distributed surveillance, and a Multi-University Research Initiative (MURI) grant focused on perception-in-the-loop abstraction, specifications, verification, learning and repair techniques that integrate perception and action of autonomous systems operating in unstructured and changing environments. The facility will located in a newly renovated robotics space at Cornell, portions of which are shared by 10 researchers.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N000141812402
Entities
People
- Mark Campbell
Organizations
- Cornell University
- Office of Naval Research
- United States Navy