Robust, Distributed Control for Coordinating Aerial Vehicles
Abstract
This research aims to design control systems that can adapt to the naval need for collaborative and autonomous formations of heterogenous vehicles. A key challenge lies in ensuring these formations can collaborate and interact despite their varied performance goals and designs. Achieving a network of autonomous agents will require a shift in perspective on control theory; we will use old theorems in a new way that allows us to focus on individual subsystems instead of the whole complicated system at once. This will allow us to adapt our tools to individual subsystems, exploiting multiple sources of information - including data, traditional models, and physics first principles. Notably, our novel, dual-model perspective will structure our controllers to combine traditional methods with neural networks. With this approach, we can create versatile tools for analysis and design that use physical information and machine learning to their fullest extents. To achieve our goals, we will develop new algorithms that analyze nonlinear dynamical models using simple and fast approximations that account for both stability and performance. These will be a steppingstone towards robust data-driven designs. More generally, our strategy of augmenting optimal designs with stability criteria allows our methods to be utilized with nearly any control design strategy. Once the basic analysis and design strategies have been developed, we will turn towards the unique challenges of networked systems, namely how their scale incurs communication and computational challenges. We will develop decentralized and distributed schemes so that only local information is needed to design and operate controllers. These schemes will be extended to leverage any available communication to improve system performance. This task requires studies on sparse communication to identify what links are most worth maintaining, as well as studies on how to add and remove agents as communication link availability changes. Finally, the problem of incorporating communication delays, which is inherent to most wireless networks, but especially those involving underwater vehicles, will be addressed. Through experimentation with networked UAVs, we will develop transformational improvements to UAV capabilities and a general control design method that is easily transferable to other systems. Our research will enhance the Navy’s capacity to control large-scale networks of autonomous vehicles and will improve naval offensive, defensive, and surveillance capabilities, allowing the navy to achieve and maintain undersea dominance and autonomous sensing
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 04, 2022
- Source ID
- N000142312043
Entities
People
- Leila Bridgeman
Organizations
- Duke University
- Office of Naval Research
- United States Navy