Agile Swarming of UAVs in GNSS-denied Plain Environments Without Explicit Communication

This paper introduces a decentralized swarm framework for the fast flight of UAVs using exclusively onboard sensory equipment. To achieve this, a bank of Linear Kalman Filters (LKFs) is designed in this work to model a UAV's neighborhood. The model improves the onboard mutual perception of UAVs and is crucial for the proposed novel flocking state feedback control rules. The control rules are designed to decrease inter-agent oscillations, which are common in standard reactive swarm models. Our swarm framework is supplemented with an enhanced Multi-Robot State Estimation (MRSE) strategy to increase the reliability of purely onboard localization that may be vulnerable to the influences of the plain environment. Although MRSE and the used LKF neighborhood model rely on information exchange between agents, we also introduce a communication-less version of the swarming framework based on estimating communicated states. The proposed solution has been verified within a set of demanding real-world experiments to demonstrate its overall capability. For motivation, we tested the swarming framework in a UAV interception task, where the swarm could follow a dynamically moving object with a group velocity overcoming 5 m/s. This paper concludes with a comparison of communication and communication-less approaches made in realistic simulations.