Generative AI-enabled Sensing and Communication Integration for Urban Air Mobility

The deepening process of urbanization poses formidable challenges to the current transportation carrying capacity. The utilization of near-ground space (NGS) and urban air mobility (UAM) greatly enhance spatial dimensions and traffic flexibility of the transportation system. However, the current limited sensing capability falls short in meeting the real-time collaborative environmental sensing and intelligent control requirements of aerial transportation. Integrated sensing and communication (ISAC) combines the sensing system of UAM with 6G communication technologies, enabling them to collaborate and achieve data sensing, transmission, processing, and decision control. The use of artificial intelligence-generated content (AIGC) facilitates real-time data fusion and decision-making, adapting to dynamic and unpredictable environments. In this paper, we first model and analyze the traffic flow in three-dimensional space, achieving knowledge embedding based on artificial potential energy field theory. Next, we design a multimodal data fusion neural network structure, which utilizes the Variational Autoencoder (VAE) to generatively achieve feature fusion and compression. Finally, we construct a UAM digital simulation platform using AirSim, which generates considerable aerial data. The simulation results demonstrate that our proposed approach achieves a feature recognition accuracy of 90.38%. The total latency is below 0.6ms, which exhibits high real-time performance.