Optimizing Cooperative Games for Cognitive Communication UAVs with Q-Learning
(Room S320C)
29 Nov 17
11:00 AM
-
11:30 AM
Currently, distributed communications networks based on multiple unmanned aerial vehicles (UAVs) are limited in terms of reliability and network availability. The capacity for each UAV to serve as a node in the network is constrained by limited energy stores, dynamic changes in the network topology, and latency/jitter issues. Typical approaches to address these challenges have focused on partitioning of the network to work around the failed nodes, but the attendant degraded communications links and lengthy network outages underscore the need for a better solution. An innovative approach based on the use of a self-forming, self-organizing, cooperative, autonomous system of distributed UAV communication nodes is being investigated. By enabling each UAV to act collectively and cooperatively, a multi-UAV network’s communication links can be made more resilient, resulting in enhanced levels of network availability and improved service quality. To achieve this, we investigated the concept of opportunistic arrays to aid in the development of a cooperative, cognitive system encompassing multiple vehicles. Based on simulations, we have also been able to demonstrate that optimal vehicle positions can be directed using decision algorithms that embody elements of game theory. In addition, by implementing a cooperative reasoning engine for system-level oversight or harmonization, we were able to ensure optimal performance of the overall system and achieve enhanced levels of service quality based on multiple measures of effectiveness.