Enhancing Cooperative Control with Machine Learning

Intelligent Cooperative Path Planning for Target Tracking

(Phase II Air Force SBIR project, subcontractor: Brigham Young University)

The goal of this project is to develop intelligent cooperative controllers for cooperative Unmanned Aircraft Systems (UAS) so that they as a team adapt to dynamically changing environment in an intelligent fashion and the system performance enhances over time. We proposed a hierarchical framework to enhance cooperative control with machine learning. We chose to address multi-UAS target tracking problem in an urban environment, where occlusions due to buildings and tunnels need to be considered. The paths of the UAS need to be carefully coordinated so that the target remains visible within the field of view of at least one of the UAS.

During Phase I, we leveraged state-of-the-art machine learning techniques to enhance the target tracking performance. We performed Monte-Carlo simulations to test our intelligent path planning scheme with different target motion models. We successfully demonstrated that the scheme has the ability to learn and use the learned information to improve the target tracking performance. In particular, we showed that our system yields superior performance over several baseline systems and a cooperative system assuming a priori kinematic model of the target. Details of our results can be found in [C13].