Joshua Smith
University of Washington
This project will develop a Deep Contact Graph Routing (Deep CGR) algorithm using real space systems provided by Astrobotic. Deep CGR uses Deep Learning methods to generalize and enhance Contact Graph Routing methods used in Disruption Tolerant Networks (DTNs). Deep Neural Networks (DNNs) will be used to predict radio channel quality (Channel State Information, CSI) as a function of robot position and time, and a joint communication / robot motion planner will be developed. We expect the planner to generate emergent strategies that combine planning over robot pose and communication system state. Hardware-in-the-loop (HIL) tests will be performed in a relevant environment (a quarry on earth) to demonstrate the feasibility of the proposed solution.
