A software-defined control approach for autonomous UAV networks
Tuesday, October 06, 2020: 12:05 PM - 1:00 PM
Networks of Unmanned Aerial Vehicles (UAVs), composed of hundreds, possibly thousands of highly mobile and wirelessly connected flying drones will play a vital role in future Internet of Things (IoT) and 5G networks. However, how to control UAV networks in an automated and scalable fashion in distributed, interference-prone, and potentially adversarial environments is still an open research problem. While networks of UAVs can certainly enable a broad range of new applications, UAV orchestration is often performed through centralized control at the core of the infrastructure or manual operations. How to design simple, elastic, and optimal control strategies for infrastructure-independent UAV networks is still a challenging and open issue. First, commercially available UAVs rely on inflexible wireless interfaces (e.g., RC or Wi-Fi), which are sensitive to spatially and temporally varying topologies, dynamic RF environments, and adversarial attacks. Consequently, even basic functionalities such as network formation and point-to-point communications are impaired by unstable channel conditions and fragile network connectivity typical of infrastructure-less aerial scenarios. Second, traditional network control schemes often rely on the assumption that the network operator is aware of real-time network state information and of low-level network infrastructure details and protocol implementations (e.g., UAVs location, network topology, spectrum availability, and modulation schemes); an assumption that often does not hold in distributed aerial networks. Finally, controlling the network behavior and flight operations in a dynamic environment requires a deep understanding of the interactions between the motion and networking functionalities at all layers of the protocol stack. In this poster we introduce a new software-defined control framework for UAV wireless networks based on distributed optimization principles. In essence, our control framework provides the Network Operator (NO) with a unified centralized abstraction of the networking and flight control functionalities. High-level control directives are then automatically decomposed and converted into distributed network control actions that are executed through programmable software-radio protocol stacks. Our system (i) constructs a network control problem representation of the directives of the NO; (ii) decomposes it into a set of distributed sub-problems; and (iii) automatically generates numerical solution algorithms to be executed at individual UAVs. We will finally present a prototype of an SDR-based, fully reconfigurable UAV network platform that implements the proposed control framework, based on which we assess the effectiveness and flexibility of our system with extensive flight experiments.
Analyst,Engineering/Technical,Research & Development,Student