Search for: All records

Deploying decentralized control strategies for outdoor multi-agent Uncrewed Aircraft Systems (UASs) is challenging due to timing variations, packet loss, and computing resource limitations. In this work we address robustness to these conditions through a novel co-regulated control strategy that varies the periodicity of control inputs and communication with other agents. Co-regulation is applied to a decentralized hierarchical controller consisting of a global component governing inter-group coordination to multiple targets while a local component governs intra-group coordination of the agents as they progress to the target of interest. The control gains are “gain scheduled” according to current conditions while a cyber controller schedules the control and communication tasks for execution based on swarm performance. The control gains are found via reinforcement learning and the entire algorithm is deployed on a swarm consisting of 7 custom agents. Our results show the impact of rethinking swarming algorithms with computation and communication resource limitations in mind and indicate we can provide exceptional swarm control utilizing fewer resources while also improving the quality of service for an onboard, anytime collision avoidance algorithm. 

Distributed multi-agent unmanned aerial systems (UAS) have the potential to be heavily utilized in environmental monitoring, especially in wetland monitoring. Deep active learning algorithms provide key tools to analyze the sensed images captured during monitoring and interpret them precisely. However, these algorithms demand significant computational resources that limit their use with distributed UAS. In this paper, we propose a novel algorithm for consensus-enabled active learning that drastically reduces the computational demand while increasing the overall model accuracy. Once each of the UAS obtains a labeled subset of images through active learning, we update the weights of the model for three epochs only on the new images to reduce the computational cost, allowing for an increased operational time. The group of UAS communicates the model weights instead of the raw data and then leverages consensus to agree on updated weights. The consensus step mitigates the impact on weights caused by the updates and generalizes the knowledge of each individual UAS to the whole system, which results in increased model accuracy. Our method achieved an average of 11.15% increase in accuracy over 25 acquisition iterations whilst utilizing only 16.8% of the processor time compared to the centralized method of active learning. 

Uncrewed Aircraft Systems (UAS) are pivotal in numerous fields, requiring dependable software architectures that reinforce functionality and e!ciency. However, e"ective in-flight monitoring of these agents is often limited to verifying hardware performance and may lack monitors for more complex software systems. The problem is seen in small UAS multi-agent systems and swarms where bandwidth is minimal and computational resources are highly constrained. Here we introduce the development, processes, and evaluation of a Health Management and Control tool tailored for monitoring the health and operational status of essential UAS software architecture components. This tool facilitates system debugging and enhances operational e!ciency through diagnostics and recovery-focused health management. 

Cyber-physical systems interact with the world through software controlling physical effectors. Carefully designed controllers, implemented as safety-critical control software, also interact with other parts of the software suite, and may be difficult to separate, verify, or maintain. Moreover, some software changes, not intended to impact control system performance, do change controller response through a variety of means including interaction with external libraries or unmodeled changes only existing in the cyber system (e.g., exception handling). As a result, identifying safety-critical control software, its boundaries with other embedded software in the system, and the way in which control software evolves could help developers isolate, test, and verify control implementation, and improve control software development. In this work we present an automated technique, based on a novel application of machine learning, to detect commits related to control software, its changes, and how the control software evolves. We leverage messages from developers (e.g., commit comments), and code changes themselves to understand how control software is refined, extended, and adapted over time. We examine three distinct, popular, real-world, safety-critical autopilots—ArduPilot, Paparazzi UAV, and LibrePilot to test our method demonstrating an effective detection rate of 0.95 for control-related code changes. 

UAVs have been studied and manufactured to help create wireless communications networks that are more flexible and cost-effective than a typical wireless network. These UAV networks could help bridge the digital divide in rural America by providing wireless communications service to areas where cell companies find it too expensive to build conventional cell towers. To test different aspects of a UAV-based millimeter-wave frequency network, we created a MATLAB simulation. The simulation visualizes a digital twin of a farm in eastern Nebraska where UAVs are tested. The simulation allows for link budgeting and interference management calculations by accommodating changes in transmitter and receiver location, frequency of the network, power of the transmitted signal, weather conditions, and antenna specifications. The simulation is able to calculate critical network values such as signal-to-interference-plus noise ratio (SINR), path loss, atmospheric loss, and antenna gains under dynamically changing conditions. 

Unmanned Aircraft Vehicle (UAV) state estimation and navigation in GPS-denied environments has received a great deal of attention, with several researchers exploring a variety of compensating estimation methods. These methods vary in capability, and usually trade off estimation accuracy for simplicity and fewer resource requirements. More advanced estimation schemes, while capable of providing good state estimates for longer periods of time, may not be suitable for small, limited resource vehicles such as UAVs. Simpler and less-accurate estimation methods, while less capable, are useful for introducing the topic to students as well as helping researchers establish flight capabilities, and may be more suitable on limited hardware. The Autonomous Vehicle Laboratory’s (AVL) REEF Estimator was designed to expedite the development of a group’s GPS-denied flight capabilities through its simple and modular design. This work seeks to extend the application of the REEF Estimator by adapting it to fit the Ardupilot flight stack so that the estimator may be used on a readily available and NDAA-compliant flight controller, specifically, a Pixhawk Cube Blue. In addition, the REEF Estimator has been containerized to further facilitate its deployment between different vehicle architectures with minimal need for reconfiguration or setup. 

Unmanned aircraft systems are expected to provide both increasingly varied functionalities and outstanding application performances, utilizing the available resources. In this paper, we explore the recent advances and challenges at the intersection of real-time computing and control and show how rethinking sampling strategies can improve performance and resource utilization. We showcase a novel design framework, cyber-physical co-regulation, which can efficiently link together computational and physical characteristics of the system, increasing robust performance and avoiding pitfalls of event-triggered sampling strategies. A comparison experiment of different sampling and control strategies was conducted and analyzed. We demonstrate that co-regulation has resource savings similar to event-triggered sampling, but maintains the robustness of traditional fixed-periodic sampling forming a compelling alternative to traditional vehicle control design.