Unmanned Aerial Vehicles (UAVs) coupled with deep learning such as Convolutional Neural Networks (CNNs) have been widely applied across numerous domains, including agriculture, smart city monitoring, and fire rescue operations, owing to their malleability and versatility. However, the computation-intensive and latency-sensitive natures of CNNs present a formidable obstacle to their deployment on resource-constrained UAVs. Some early studies have explored a hybrid approach that dynamically switches between lightweight and complex models to balance accuracy and latency. However, they often overlook scenarios involving multiple concurrent CNN streams, where competition for resources between streams can substantially impact latency and overall system performance. In this paper, we first investigate the deployment of both lightweight and complex models for multiple CNN streams in UAV swarm. Specifically, we formulate an optimization problem to minimize the total latency across multiple CNN streams, under the constraints on UAV memory and the accuracy requirement of each stream. To address this problem, we propose an algorithm called Adaptive Model Switching of collaborative inference for Multi-CNN streams (AMSM) to identify the inference strategy with a low latency. Simulation results demonstrate that the proposed AMSM algorithm consistently achieves the lowest latency while meeting the accuracy requirements compared to benchmark algorithms.

UAV networks often encounter jamming attacks, under which multi-radio protocols have to switch radios to accelerate communication recovery. However, the existing protocols rely on exchange of hello messages to detect jamming, leading to long sensing time and thus slow routing recovery. To address the issues raised by jamming attacks, we propose a new routing protocol, Electromagnetic Spectrum situation awareness Optimized Link State Routing (ESOLSR) protocol, to improve the existing OLSRv2 protocol. ESOLSR utilizes the spectrum situation awareness capability from the physical layer, and adopts joint-updating of link status, updating of interface functions, and adaptive adjustment of parameters. Our simulation results show that the improved protocol, ESOLSR, can recover routing and resume normal communication 26.6% faster compared to the existing protocols.

Low-Altitude Intelligent Network (LAIN), as a new type of intelligent network, relies on space-air-ground-sea facilities to constitute a digital intelligent network system. It is a key component of the space-air-ground integrated network, and can support the seamless and ubiquitous connections of the sixth generation communication technology and promote the development of intelligent network service from ground to low-altitude space. However, LAIN is still in the developing stage and faces the following key challenges: intractability of aerial control, severe spectrum interference, and multi-dimensional resource limitation. This article focuses on the issues of LAIN architecture and safety control, including the current development status of low-altitude network and its significance for industrial technology transformation. Then, from the perspective of spectrum resources, network resources, and airspace resource management, the recent related works are analyzed. Furthermore, we analyze the key technologies such as low-altitude aircraft air-ground spectrum sharing, sensing, transmission, computing networking coverage, low-altitude airspace intelligent supervision, and point out future development directions. Finally, an application demonstration of LAIN is proposed, aiming to satisfy the significant requirements for efficient operation and safety in low-altitude airspace, and providing the theoretical basis as well as technology for the further development of the next generation space-air-ground integrated network.